--- updatedAt: 2026-09-01T22:35:04.000Z agentTools: projectIndex: https://documentation.picsellia.com/llms.txt --- # Prediction Review Campaign An *Prediction Review Campaign* allows you to handle properly and efficiently the process of reviewing the `PredictedAsset` of a `Deployment` on Picsellia. By using *Prediction Review Campaign* on Picsellia, you will be able to: * Define your *Prediction Review Workflow* * Assign *Review & Quality Check Tasks* * Check the *Reviews and Quality Checks* done and raise *Issues* if something is wrong * Track the progress of the *Prediction Review Campaign* ## 1. Create an *Prediction Review Campaign* First of all, it is important to note that each `Deployment` can have up to one *Prediction Review Campaign*. To initialize an *Prediction Review Campaign* on a `Deployment`, go to the *Settings* tab, and click on *Workflow* in the *Campaign* sub-section. From the *Prediction Review Workflow* creation view, you can choose either to create an *Empty Workflow* (i.e. without *Steps* already created) or a *Default Workflow* (i.e. a basic two *Steps* *Workflow* that you can for sure modify).

Prediction Review Workflow creation ### A. Create an *Annotation Workflow* Let's consider that you are starting from an *Empty Workflow*. Here, you will be able to define the *Prediction Review Workflow* that defines the *Steps* your `PredictedAsset` will follow during the *Prediction Review Campaign* you are about to launch on this `Deployment`. The *Prediction Review Workflow* represents all the *Steps* the `PredictedAsset` will go through before being considered properly reviewed. As you can see above, the *Workflow* is initially free of *Steps*, so you will now be able to define and configure new *Steps* before launching the *Prediction Review Campaign*. To do so, we just need to click on the *+* button next to the initial step of your *Prediction Review Workflow*. A new *Step* will be added to your *Prediction Review Workflow*, it is now time to configure it.

Create a new step in your Prediction Review Workflow Once the new *Step* is created, its related configuration panel will pop up on the right side. First, you'll be requested to give a name to your *Step* and to define if it is a *Review* or a *Quality Check* *Step*. > 📘 The only difference between a *Review* and a *Quality Check Step* is the ability for *Quality Check Step* to Accept or Reject what has been done in previous steps.

Step type configuration Then, you need to select the *Assignees* that will get *Tasks* attributed as part of this *Step*. As shown below you can select the *Assignees*.

Step Assignee configuration > 📘 Select *Assignees* > > The potential *Assignees* are the members of the current *Organization* that are listed are *Collaborators* in the `Deployment`. > > So if a member is not part of the *Assignee* list, you should add him as *Collaborator* in the related `Deployment`. Finally, you'll be requested to define the assignation strategy based on two criteria. * The *Asset split strategy* allows you to define if the `PredictedAsset` going through this *Step* should be equally assigned to all the selected *Collaborators* (*Equally dispatched* strategy) or if some *Collaborators* should get more *Tasks* assigned then others. In this case, you could set it up using the *Custom Weights* strategy. For *Custom Weights* strategy the percentage of *Tasks* assigned to each *Collaborator* will be displayed as shown below. * The *Sample rate*, allowing you to set the probability an image submitted at the previous step has to actually go through the current *Step*. This one could be really useful in case of want to perform a *Prediciton Review* on only a sample of the annotated images.

PredictedAsset split strategy & sample rate definition Now that your *Step* is fully configured you just need to click on *Create* to finish the *Step* creation. Once created, you can at any time find out the number of *Assignees* assigned to this *Step* and the number you can edit the step anytime by clicking on it and accessing the configuration panel.

Update the configuration of an existing step > 👍 The *Unpriviledged* role > > If you want some members to only have access to the *Tasks* they are supposed to do within *Prediction Review Campaign* running in your *Organization* and nothing else, you can set them as *Unpriviledged* members. More details are available [here](https://documentation.picsellia.com/docs/campaign#6-the-labeler-role). Then, you can add as many *Steps* as you want to create the *Prediction Review Workflow* that suits your needs. For instance, a simple and efficient *Prediction Review Workflow* would be to have an *Review* *Step* and then a *Quality Check* *Step* with a part of your team assigned to *Review* and some others to the *Quality Check*, as shown below:

Example of a simple Prediction Review Worfklow > 📘 A *Collaborator* can be assigned to several *Steps* of an *Prediction Review Workflow* > > In this case, the *Tasks* will be assigned to avoid a given `PredictedAsset` to be reviewed and quality checked by the same *Collaborator* when possible. ### B. Configure your *Prediction Review Campaign* Now that your *Prediction Review Workflow* is defined, you can configure the way your *Prediction Review Campaign* will be initialized by clicking on the settings button. A configuration panel will open allowing you to: * Add an instruction PDF file for the *Collaborators* which will be visualizable by all the *Assignees* of the current *Prediction Review Campaign* while doing the *Tasks* they get assigned to. This pdf could, for instance, list all the guidelines and instructions the *Collaborators* need to be aware of for properly reviewing the `Prediction`. * Define a due date for the *Prediction Review Campaign* * Make sure the *Prediction Review Campaign* is closed when all the *Tasks* are Done

Workflow Settings ### C. Launch the *Prediction Review Campaign* At the end of the configuration phase of your *Prediction Review Campaign*, all the `PredictedAsset` and associated `Prediciton` of your `Deployment` are at the initial *Step* named *Deployment*.

PredcitedAsset at the initial step It is now time to start the *Prediction Review Campaign*. To do so you simply need to click on *Sync Manually*, this will make all the `PredictedAsset` present in the *Deployment Step* transition to the first *Step* of the *Prediction Review Workflow*, the related *Tasks* will then be created.

PredictedAsset at the first Review step Now that the *Campaign* is live, the *Campaign* badge will be displayed and two new tabs will be created, *Tasks* and *Issues*. They will allow you and the other *Organization* members to go through this ongoing *Prediction Review Campaign*.

Campaign launched ## 2. Access the *Tasks list* As soon as an *Prediction Review Campaign* is ongoing on a `Deployment`, you will be able to access a new tab called *Tasks*. > 🚧 Who can access the *Tasks* tab ? > > This *Tasks* tab is available for every *Collaborator* in your `Deployment` except the ones with the `Labeler`role (those added as *Unprivileged* to your *Organization*). > > The members with the role *Unpriviledged* in the *Organization*, that have been added in the *Collaborators* list of the `Deployment` can only visualize the *Tasks* that are assigned to them in a dedicated view as detailed [here](https://documentation.picsellia.com/docs/campaign#6-the-labeler-role). The *Tasks list* allows you to visualize the current *Status* and *Step* of the latest *Tasks* related to any `PredictedAsset` that is currently going through the *Prediction Review Campaign*.

Tasks list This view allows you to filter the *Tasks* to display. You can filter by the *Task* *Status*, by the current *Step* of the *Prediction Review Workflow*, or by the current *Assignee*.

Filter Tasks ### A. Task Status As shown below each *Task* created for an `PredictedAsset` has a current *Step* and *Assignee*, those are defined while creating the *Prediction Review Workflow* as described [here](https://documentation.picsellia.com/docs/campaign#b-create-an-annotation-workflow). However, we can also notice the presence of a new field named *Status* for each *Task*. This *Status* field describes the status of the `PredcitedAsset` journey in the *Prediction Review Workflow*. The *Status* of a *Task* can either be: * *In Progress*: Everything is going well until now, the `PredictedAsset` and its associated `Prediction` are going through the *Prediction Review Workflow* normally. However, the `PredictedAsset` is still at a *Review* or a *Quality Check* step reflected by the *In Progress* status. The *Prediction Review Workflow* isn't over yet. * *To Rework*: A *Task* with the *to Rework* status is a *Review* *Task* created when a `PredictedAsset` has been annotated during a first *Review Task* and submitted as *To Rework* during the next *Quality Check* step. Indeed, the *Assignee* of this *Quality Review* step didn't accept the proposed *Review*. So he sent the `PredcitedAsset` back to the previous *Step* in the *Prediction Review Workflow*. Usually, the reason is described in an [Issue](https://documentation.picsellia.com/docs/campaign#5-issue-management). As a consequence, a new *Task* with *To Rework* status has been re-created at the previous *Review* step of the *Prediction Review Workflow* and re-assigned to the *Collaborator* that created the rejected *Review*. This *Assignee* can then modify its first *Review* by completing this *To Rework Task* and sending it back to the following *Quality Check* *Step*. * *Done*: The `PredictedAsset` and its associated `Prediciton` has been through the whole *Prediction Review Workflow*, the `PredictedAsset` can be considered as reviewed. It will be displayed with the status *Reviewed* in the *Predictions* tab and as a consequence can be submitted to the *Feedback Loop*. * *Skipped*: While completing an *Review* or *Quality Check Task*, the *Assignee* has the possibility to *Skip* the *Task*, this will kick the associated `PredictedAsset` out of the *Annotation Workflow*. There is no *Task* to be completed on this `PredictedAsset` anymore but it cannot be considered as properly reviewed. Usually, the reason is described in an [Issue](https://documentation.picsellia.com/docs/campaign#5-issue-management). ### B. Reassign a *Task* or a bunch of *Tasks* As described in detail [here](https://documentation.picsellia.com/docs/campaign#3-complete-a-task), a *Task* can only be completed by the *Assignee*. If for some reason, a *Task* needs to be completed by someone else, this one can be reassigned for completion to a member that is identified as *Collaborator* on the `Deployment`. If you have sufficient rights, it can be done for a particular *Task*, by clicking from the *Task list* on the *Assignee*, the list of available *Collaborators* will then be displayed. You just need to click on the new *Assignee* as shown below:

Reassign a Task If you have sufficient rights, it can be done for a bunch of *Tasks*, by selecting the to-be-reassigned *Tasks* and clicking on the *Reassign* button. Then you simply need to select from the list of available *Collaborators* the one that will get the selected *Tasks* reassigned.

Reassign Tasks > 🚧 A filter on the *Steps* should be active > > For the reassignment of several *Tasks*, you must ensure that a filter on the *Task* step is active on the *Task list* before reassigning. ### C. Submit a bunch of *Tasks* During a normal functioning process, each *Task* should be completed by its *Assignee* by opening the *Prediction Review Tool*. But sometimes, *Collaborators* prefer to first complete their *Tasks* in the *Annotation Studio*, save them, and *Submit* all their *Tasks* at the same time. In this case, they can use the *Submit* button available in the *Task list*. To *bulk submit* a bunch of *Tasks*, you simply need to use the filters or the *Query Langage* to retrieve and select *Tasks* to be submitted and click on *Submit*:

Bulk Submit By clicking on *Submit*, all the selected *Tasks* will be submitted, meaning that the `PredictedAsset` will transition through the *Annotation Workflow* to the next *Step*. A new *Task* will be created according to the *Task* assignation rules of the *Step*. Please be aware that only the selected `PredcitedAsset` with an existing *Review* will be submitted. If you want to submit also the`PredcitedAsset` without *Review* you can use the *Move* button described [here](https://documentation.picsellia.com/docs/campaign#d-move-a-bunch-of-tasks). In addition, if the next *Step* in the Workflow is *Done*, the `PredcitedAsset` with *Issues* still open won't be submitted. Once again, if you want to submit also the`PredictedAsset` open *Issues* related, you can use the *Move* button described [here](https://documentation.picsellia.com/docs/campaign#d-move-a-bunch-of-tasks). ### D. Move a bunch of *Tasks* During a normal functioning process, each *Task* should be completed by its *Assignee* by opening the *Annotation Studio*. But in some cases, you might be forced to move several *Tasks* at the same time to another *Step* (including *Done* or *Skipped*), whether they are assigned to you or not. To do so, you simply need to use the filters or the *Query Langage* to retrieve and select *Tasks* to be transitioned to another *Step*. Then click on *Move*:

Bulk Move A pop-up will then open allowing you to select the destination *Step* of the selected *Tasks*:

Select Destination Step Once `PredictedAsset` are move to another *Step*, obviously new *Tasks* will be generated at this new *Step* according the *Task* assignation rules of this one. > 📘 `PredictedAsset` moving backward in the *Workflow* > > If the selected `PredictedAsset` are sent backward in the *Prediction Review Workflow* the new *Task* created will have the *In Rework* status, whereas if `PredictedAsset` are more forward in the *Prediction Review Workflow*, the new *Task* created will have the *In Progress* status. Please be also aware that unlike the *Submit* button, using the *Move* button will transition the selected `PredictedAsset` even if some *Issues* are still open or if no *Review* has been initiated on this `PredictedAsset` yet. The *Move* operation can also be done from the *Worflow Settings* as shown below. However, from this tab, you cannot granularly select which *Tasks* from a given step should be moved or not.

Move from Workflow Settings ### E. Redistribute Tasks At any moment, the distribution of the *Tasks* at each *Step* may have changed from the one defined at the *Step* definition. This can be explained by the manual reassignation of *Tasks*, by some *Assignees* that are faster than others or by the update of custom weights. For all those reasons, you have the possibility for each *Step* to redistribute among the Assignees the current tasks according the the latest *Weights* defined , to do so you just need to click on *Redistribute* as shown below:

Redistribute the Tasks of a Step aming the Assignees according to the defined weights As you can see above, for each *Step* selected, you have an overview of the real and theoric (defined in the *Assets split strategy*) Tasks repartition among Assignees. ## 3. Complete a *Task* Now that we can visualize, understand, and manipulate the *Tasks* created in the frame of the current *Prediction Review Campaign*, it is high time to start completing them to make our `PredictedAsset` go through the *Prediction Review Workflow*. > 📘 Who can complete a *Task*? > > The only member able to complete a *Task* is the *Assignee*.\ > So if you want to complete a *Task* without being the *Assignee*, you need to reassign it first as explained [here](https://documentation.picsellia.com/docs/campaign#b-reassign-a-task-or-a-bunch-of-tasks). Whatever the *Task* type (*Annotation* or *Review*) to complete it you need to click on the *Go to task* button as shown below, the *Prediction Review Tool* will then open allowing you to perform the *Annotation* or the *Review* required by the *Task*. The `PredictedAsset` can be free of `Annotation` or in case the `Annotation` has already been initialized this one will be displayed as submitted at the previous step of the *Annotation Workflow*.

Complete a Task ### A. Complete an *Review Task* In case you need to complete a *Review Task* the *Prediction Review Tool* will open displaying the `PredictedAsset` to be reviewed. You can rely on the documentation of the *Prediction Review Tool* [here](https://documentation.picsellia.com/docs/deployments-predictions-review#5-review-a-prediction) to create the `Shape` that will compose the *Review*. Once you have created or edited the `Shape` and considered the *Review* satisfying, you can *Submit* the *Task* as shown below.

Submit a Task Submitting a *Task* will save it and transition it to the next step of the *Prediction Review Workflow*.\ Please note that you can also just *Save* the *Review* and submit the *Task* later on. However, as you are reviewing within a *Prediction Review Campaign*, the *Prediction Review Tool* has some additional features dedicated to *Prediction Review Campaign*. Please note that any *Collaborators* who’s not a *Labeler* and who has access to the `Deployment` can click on the *Go to task* button and open the related `Asset` in the *Annotation Tool*, but only the *Assignee* of the *Task* can create, edit the *Review* and submit the *Task*. #### a. Accessing the history of the `PredictedAsset` From the *Prediction Review Tool* you can visualize the history of the `PredictedAsset` during the *Prediction Review Workflow* as shown below:

Workflow History You can see all the steps done or to come of the *Prediction Review Workflow*. The current step is highlighted in orange. For each step are displayed, the *Assignee* and the submission date. In addition by clicking on any step, you can visualize the *Review* submitted at this step allowing you to see and track how it evolved going through the *Prediction Review Workflow* . #### b. Creating *Issues* To ease the communication between the different *Collaborators* involved at the different steps of the *Workflow*, *Issues* can be raised on the whole `PredictedAsset` or a zone of the `PredictedAsset`. *Issues* can for instance be used when a *Collaborator* doubts how to annotate an image to request advice from the *Collaborator* in charge of the *Quality Check*. It can also be used to track the justification why a `PredictedAsset` has been skipped and sent back to an *Review step*. More details on the usage of *Issues* during an *Prediction Review Campaign* are available [here](https://documentation.picsellia.com/docs/campaign#5-issue-management). #### c. Visualizing the *Annotation Instructions* As explained [here](https://documentation.picsellia.com/docs/campaign#1-create-an-annotation-campaign), during the *Prediction Review Campaign* initialization, a pdf file can be attached.\ This can be used to share the annotation guidelines with all the *Collaborators* who will be involved in the *Prediction Review Campaign*. During the completion of a *Task*, a *Collaborator* can open the pdf file as shown below:

Campaign Instructions ### B. Complete a *Quality Check Task* In case you need to complete a *Quality Check Task*, the *Prediction Review Tool* will open displaying the `PredictedAsset`, the related *Review* and the potential *Issues* created previously during the *Prediciton Review Workflow*. During a *Quality Check Task*, the *Assignee* has to review the *Review* done during the previous *Review step*. During a *Quality Check Task*, 4 scenarios can occur: * The *Review* is correct according to the *Assignee*, the *Task* can be submitted as it is by clicking on the *Submit* button. * The *Review* is not satisfying and the *Assignee* wants to edit it himself. The *Assignee* can then use the *Prediction Review Tool* to edit the *Review*. Once it is considered as satisfactory, the *Assignee* can click on *Submit* to transition the `PredictedAsset` to the potential next step of the *Workflow*. * The *Review* is not satisfying and the *Assignee* wants to send the `PredictedAsset` back to the previous *Review step* allowing the previous *Assignee* to rework the *Review*. Before sending it back, the *Assignee* will have to create an *Issue* (*Zone or Image Issue*) to explain why the `PredictedAsset` is sent back to the previous step, once done, the *Assignee* of the *Quality Check task* can click on *To Rework,* and a new *Task* will be created and assigned at the previous step of the *Workflow*. * The *Assignee* of the *Quality Check Task* considers that the `PredictedAsset` should not keep going through the *Prediction Review Workflow*. In this case, the *Assignee* can click on *Skip*, then the *Assignee* will be requested to create an *Issue* to explain why this one should be skipped. During the *Quality Check step* while completing its *Task*, the *Assignee* can answer or close the *Issues* created previously by other *Collaborators* and create new *Issues* that will be handled by the next *Assignee*. (either further in the *Workflow* if the `PredictedAsset` is *Submitted* or upstream if the `PredictedAsset` is sent to rework). ## 4. Manage *Prediction Review Workflow* lifecycle At any moment, you can update, close or delete your *Prediction Review Campaign*. ### A. Update a *Prediction Review Campaign* As explained previously you can manage easily manage your Tasks by transitioning, submitting, or reassigning them. #### a. Update a Step But you can also during a *Prediction Review Campaign* modify your *Workflow*. Each step configuration can be updated at any moment by modifying its *name*, *Sample rate*, *Assignees*, or *Asset split strategy*.

Access the Step editing panel

Step editing panel #### b. Add a Step You can also add a *Step* at any moment of an *Prediction Review Campaign* by clicking on the *+* at the desired place for your new *Step*.

Add a new Step The *Step* configuration panel will then open allowing you to configure this new step the same you already did during *Worflow* initialization.

New Step configuration As soon as you click on *Create*, the new *Step* will enter the Workflow. This meaning that submitting a `PredictedAsset` at the previous *Step* will make it go through it. Additionnaly, any `PredictedAsset` can be moved to this new *Step*. #### c. Delete a step You can also delete a *Step* at any moment of a *Prediction Review Campaign* by clicking on the *Step* and *Delete*.

Step Deletion In case, there are currently no *Tasks* at the *Step* to be deleted, then no problems, the *Step* is actually deleted. However, if there are currently *Tasks* at this *Step*, a pop-up will open asking you to select the *Step* those *Tasks* should be sent to.

Task management of a to-be-deleted Step This way you can delete a *Step* while ensuring there is no impact on its related *Tasks*. You can anyway always display the deleted *Tasks* using the radio button as shown below:

Show deleted Steps For traceability purposes, even if a *Step* is deleted, all the work done at this one when it was alive is kept and can be visualized in the *History* tab of the *Prediction Review Tool*.

Asset history with a deleted Step This way you can delete *Steps* of your *Prediciton Review Workflow* without losing any information ### B. Delete an *Prediction Review Campaign* A *Collaborator* with *Admin* right on the parent `Deployment` can decide to *Delete* an ongoing *Prediction Review Campaign*. It can be made from the *Deployment Settings* page, through the *Campaign settings* button as shown below:

Prediction Review Campaign Deletion The deletion of a *Prediction Review Campaign* is irreversible and for each existing `PredictedAsset`, only the latest version of the related *Review* will be kept. In other, words you will lose all the history of the *Prediction Review Campaign* and won't be able to access the *Tasks* & *Issues* tabs anymore. Therefore we strongly recommend you to be very cautious while deleting an *Prediction Review Campaign*. Once the *Prediction Review Campaign* is deleted, the `Deployment` is back to its initial state but each `PredictedAsset` still has the latest version of the *Review* created through the deleted *Prediction Review Campaign*. For sure, from this point you can create brand new *Prediction Review Campaign*. ## 5. *Issues* Management As already mentioned previously, within a *Prediction Review Campaign*, any *Assignee* can create an *Issue* while completing its *Task*. The *Issue system* aims at facilitating communication between the different *Assignees* for a given `PredictedAsset` along its transition through the *Annotation Workflow*. Once an *Issue* is created the different *Collaborators* can chat and comment on this one until it is closed. All the open and closed *Issues* related to a `PredictedAsset` are accessible by authorized *Collaborators* at any step of the *Workflow*. Please note that the *Issue system* is only available for `Deployment` with a *Prediction Review Campaign* ongoing. There are two types of *Issues*: * *Zone issue*: Related to a specific area of the `PredictedAsset` * *Image Issue*: Related to the whole `PredictedAsset` ### A. Create a *Zone Issue* This *Issue* can be related to the whole `PredictedAsset` or to a specific zone on the `PredictedAsset`. When completing a *Task* on the *Prediction Review tool*, as an *Assignee* you can create a *Zone Issue* by clicking on the *Create a zone Issue* button as shown below. Then draw on the image a rectangle on the zone affected by the *Issue* you want to raise. Please note that any *Collaborator* who can access the *Task* (even without being the *Assignee*) can create a *Zone Issue* on the `PredictedAsset`.

Create a Zone Issue Once the zone has been created, the *Issues Panel* will open in the right panel of the *Prediction Review tool*. You can then describe your *Zone Issue* and click on *Create Issue*.

Zone Issue creation Once created, your *Issue* will appear on the *Issues panel* related to the current `PredictedAsset`.

Issues panel ### B. Create an *Image Issue* An *Image Issue* is not related to a specific area but to the whole `PredictedAsset`. An *Image Issue* can be created directly from the *Issues Panel* as shown below:

Image Issue creation Once the description is prompted the *Image Issue* will be created after clicking on *Create Issue*. This new *Issue* will be also accessible at any time from the *Issues panel* .

Issues panel Please note also that the *Assignee* will be forced to create an *Image Issue* before *Skipping* or sending a `PredictedAsset` for *Rework* if nor *Image or Zone Issue* exists on the current `PredictedAsset`. This is a way to track why this `PredictedAsset` has been *Skipped* or sent *To Rework*. As it is the case for *Zone issues*, from this *Issue Panel*, *Collaborators* can comment, resolve or re-open the *Image Issue*. ### C. Issues panel After opening any `PredictedAsset` from the *Tasks list*, you can access the *Issues panel* as shown below on the right side of the *Prediction Review Tool*.

Issues Panel Here are displayed all the *Issues* (Zone or Image) related to the current `PredictedAsset` created during the *Prediction Review Campaign*. You can switch between the still-open *Issues* and the closed ones. By clicking on any *Issue* (Open or Closed), you will access the conversation thread allowing you to see the conversation related to this issue, add a Reply, Resolve it, or Re-Open it.

Reply and/or Resolve an Issue

Reply and/or Resolve an Issue Please note that any *Collaborator* that can access the *Task* (even without being the *Assignee*) can interact with an *Issue* in the *Issue Panel*. Please note that any *Issue* opened will remain until a *Collaborator* with the necessary rights closes it. > 👍 An `PredictedAsset` cannot turn in *Done* status with still open *Issues* > > To make a `PredictedAsset` turn in the *Done* status, meaning submitting the `PredictedAsset` at the last step of the *Workflow*, all the *Issues* related should be closed. In a few words, you won't be able to submit the very last *Task* of the *Workflow* is some *Issues* are still open. ## 6. The *Labeller* role In case you need to involve some members in your *Prediction Review Campaign* but willing them only to complete the *Tasks* they get attributed without letting them visualize or manipulate any other object related to your *Organization*, you can add them as *Unpriviliedged* and add them with the *Labeler* role in the `Deployment`. ### a. Add a member as *Labeler* in your `Deployment` To involve a member in your *Prediction Review Campaign* and let him complete the assigned *Tasks* without accessing any other object of your *Organization*, you first need to add this member with the *Unpriviledged* role to your *Organization*. To do so, you need to be *Admin* in the current *Organization*, go to *Organization Settings > Members* and click on *Inviter member*.

Access members panel of an Organization You can add members using their username if they already have a Picsellia account or by typing their email address (in this case, they will get an email with a link to create their account), then you need to select *Unpriviliedged* as *Default Permission* and click on *Invite*.

Invite an Unprivileged member Then to make sure you can involve the invited *Unprivileged member* in an *Prediction Review Campaign*, you need to add him to the *Collaborators list* of the `Deployment` as shown below:

`Deployment` Collaborators list Once done the *Unprivildged member* is identified as *Labeler* in the `Deployment` and can be involved in *Prediction Review Campaign* for any `Deployment` related. ### B. Access an *Organization* as an *Unprivileged* member The members invited to an *Organization* with *Unprivilideged* rights, can access the *Organization* but by default, they won’t be able to access any object (`Datalake`, `Dataset`, `Project`, `Model`, `Deployment`…). They will only be able to visualize the *My Tasks* view. This view is only accessible to members with the *Unpriviledged* role in the current *Organization*. From this one, the *Unpriviledged* member will see the list of *Tasks* assigned to him and nothing else.

My Task view for Labeler The *Labeler* will see all the *Tasks* assigned within the current *Organization* grouped by *Prediction Review Campaign*. Each *Prediction Review Campaign* is identified by its `Deployment` related. The *Labeler* can visualize all the *Tasks* that are currently assigned to him grouped by *Prediction Review Campaign*, the name of the current step that generated this *Task*, and the number of issues on the related `PredictedAsset`. The *Labeler* can also filter on the *Tasks Status* or *Step* name. To complete a *Task* as *Labeler*, you can click on *Go to task* and you can complete your *Task* (*Review* or *Quality Check*) as detailed [here](https://documentation.picsellia.com/docs/campaign#3-complete-a-task). The *Labeler* can also access a detailed view describing all the *Issues* raised on any `PredictedAsset` that generates a *Task* currently assigned to the *Labeler*.

Issues list for a Labeler > 📘 *Labelers* can manage *Issues*, read instructions and access `PredictedAsset` history > > For any `PredictedAsset` that generate a *Task* currently assigned to a given *Labeler*, this one can access all the *Issues* management features (Open, Reply, Resolve, Re-open). > > While completing a *Task* the *Labeler* can also access the history of the `PredictedAsset` int he *Prediction Review Workflow* and open the *Review guidelines* file if any. ## 7. Analytics *Collaborators* who are not *Labelers* can access the *Analytics Dashboard* related to the current `Deployment` in the *Analytics* tab. If an *Prediction Review Campaign* has been set, the three following tabs are available: * *Throughput*: This tab will through its metrics detail the progress of the *Prediction Review Campaign*. * *Efficiency*: This tab will through its metrics detail the time spent by the *Assignees* on the *Prediction Review Campaign.* * *Quality*: This tab will detail the `Deployment` with statistical metrics based on the *Review* created until now. > 📘 Only *Quality Analytics* are availble out of an *Prediction Review Campaign*.

Access Analytics and associated tabs ### A. Throughput The *Throughput* tab displays graphs that are related to the progress of the *Prediction Review Campaign*. From this tab, you will have a clear view of how the `PredictedAsset` of the `Deployment` is going through the *Prediction Review Workflow* in real-time. In detail, the following metrics are displayed: * *Prediction Review Campaign Progress*: A percentage of *Prediction Review Campaign* completion based on the *Tasks* already submitted compared to the total number of *Tasks* to submit to complete the *Prediction Review Campaign* * *Tasks Distribution*: A pie chart showing the total number of active *Tasks* grouped by *step*. * *Total Review Assignments*: A bar chart showing the total number of *Annotation Tasks* submitted. The time range depends on the selected filter. * *Total Quality Check Assignments*: A bar chart showing the total number of *QUality Check Tasks* submitted. The time range depends on the selected filter. * *Assignment Count By Collaborator*: A bar chart detailing the total number of *Tasks* currently to each *Collaborator* involved in the *Prediction Review Campaign*. * *Issues Quantity*: A bar chart detailing the total number of Issues created until now during the *Prediction Review Campaign* grouped by *Creation Date*. The time range depends on the selected filter as explained below. * *Issues Type Distribution*: A pie chart detailing the total number of Issues created until now during the *Prediction Review Campaign* grouped by *Issues Type*(Zone or Image). The time range depends on the selected filter.

Throughput Analytics For each chart, you can change the *Date Range* of the data displayed as shown below:

Date Range In addition, you can filter the data displayed based on a given step of the *Prediction Review Workflow* or by *Assignee*.

Filter by Assignee

Filter by Assignee Filter by step or Assignee

Filter by step ### B. Efficiency The *Efficiency* tab displays graphs that are related to the time dedicated to the different steps of the *Prediction Review Campaign* in real-time. In detail, the following metrics are displayed: * *Task Duration*: Globalized time information on all the *Tasks*, the average time to complete a *Task*, and the minimum and maximum time of *Task* completion on the whole current *Prediction Review Campaign*. * *Closed Issues*: The total number of *Issues* closed during the current *Prediction Review Campaign*. * *Opened Issues*: The total number of *Issues* opened during the current *Prediction Review Campaign* * *Total Review Time*: The total time spent on *Review Tasks* until now. A bar chart displays this duration on the selected Date Range based on the filter. * *Average Review Time*: The average time spent on one *Review Task* until now. A bar chart displays this duration on the selected Date Range based on the filter. * *Total Quality Check Time*: The total time spent on *Quality Check Tasks* until now. A bar chart displays this duration on the selected Date Range based on the filter. * *Average Quality Check Time*: The average time spent on one *Quality Check Task* until now. A bar chart displays this duration on the selected Date Range based on the filter.

Efficiency analysis For each chart, you can change the *Date Range* of the data displayed as shown below:

Date Range In addition, you can filter the data displayed based on a given step of the *Prediction Review Workflow* or by *Assignee*.

Filter by step

Filter by Assigne