List view
## Revised SynANNO Roadmap Outline ### Phase 1: Data Preparation for Model Training **Objective:** Prepare a comprehensive dataset that will be used to develop and train the 3D UNet model for accurate auto-segmentation. **Steps:** 1. **Data Selection:** - Run queries on the materialization table to select instances with diverse characteristics in terms of slice numbers. 2. **Data Extraction and Processing:** - Cut out the required subvolumes from the raw and target volumes for the selected instances. - Adjust the target volumes to represent a realistic variety of seed segmentation scenarios. #### Phase 2: Model Training and Preliminary Integration (See Issue #70) **Objective:** Train the 3D UNet model with the prepared dataset and perform preliminary integration into SynANNO. **Steps:** 1. **Training the 3D UNet Model:** - Employ the curated dataset to train the 3D UNet, ensuring it can handle variations in the number of slices and achives satisfying segmentaiton results. 2. **Model Validation:** - Validate the model's performance using a separate validation set extracted in a similar manner as the training set. #### Phase 3: Enhancing User Interaction and Control **Objective:** Build upon the existing UI to add user-initiated control for auto-segmentation and improve the manual review process. **Steps:** 1. **"Start Auto-Segmentation" Button:** - Integrate a "Start Auto-Segmentation" button to allow users to initiate the segmentation process after drawing a custom mask. 2. **Segmentation Review Workflow:** - Implement the UI logic to display auto-generated segmentations in subsequent slices, allowing users to review and, if necessary, draw additional custom masks. 3. **Iterative Segmentation Improvement:** - Enable the auto-segmentation to be re-triggered after additional custom masks are drawn, allowing for iterative refinement of the segmentation results. #### Phase 4: Debugging, Optimization, and User Experience (See Issue #71) **Objective:** Address known issues, optimize performance, and refine the user interface for better user experience. **Problem Areas and Solutions:** - **Processing Button Accessibility:** Implement a readily accessible processing button throughout the UI. - **UI/UX Redesign of Loading Page:** Enhance the layout and logic of the "Loading Data" page. - **Robust Error Handling:** Introduce advanced error handling for a seamless user experience. #### Phase 5: User Acceptance Testing and Feedback Integration **Objective:** Validate the developed functionalities with end-users to gather feedback and prepare for eventual deployment. **Steps:** 1. **Deploy Beta Version:** - Release a beta version of SynANNO to a select group of users for testing purposes. 2. **Collect User Feedback:** - Establish channels for collecting detailed feedback from beta testers. 3. **Incorporate Feedback:** - Analyze feedback and make necessary adjustments to the system. 4. **Pre-Production Review:** - Conduct a thorough review to ensure the system is ready for a production environment considering all user feedback.
Overdue by 2 year(s)•Due by December 31, 2023•2/3 issues closedMaking SynANNO publication-ready by making its workflows scaleable, adapting them to expert feedback, and making SynANNO CAVE integration ready. Tasks: - Collect expert user input - Enable a view-centric approach - Neuron-centric approach - Handle corner cases - Reevaluate line thickness - NG view in draw mode - Review model based depth wise auto segmentation - Emulating neuron-synapse association
Overdue by 2 year(s)•Due by August 1, 2023•18/18 issues closedWe want to roll out the first version to our collaborators at the end of this month. The work for this milestone, therefore, includes: - Debugging - Documentation - Tutorials
Overdue by 3 year(s)•Due by October 31, 2022•6/6 issues closed# Concept The view lets the user annotate the center slice of detected synapses. Our models will then calculate the missing annotations in the backend and return the full annotated target. # Views and Flows ### Landing page The first page shows two buttons. One states "Annotate," and the other states "Proof Read." Clicking on "Proof Read" takes the user to the already implemented proofreading tool (1). Clicking on "Annotate" takes the user to the annotation tool (2). ### Annotation tool ##### Uploading Clicking on "Annotate" takes the user first to a similar data loading page as in (1). However, the user will only be asked to provide the source file. ##### Annotation: The Annotation view is the same as in (2). However, it does not depict any targets, and when the user right-clicks an instance, the opening module also has a "draw annotation" mode. ###### Draw Annoation mode If the user clicks on the "Annotation" button, within the single instance module view, the module depicts the center slice of the current instances and opens a translucent drawing field above the depicted slice. The user can then set a start and an endpoint of a line marking the cleft and select one side of the line that indicates the positive polarity. Additionally, the user can drag 1-2 waypoints on the line marking the cleft to adjust its shape. If the user is happy with the result, he can save the drawn maks, which from then on get depicted for the particular instance. The user can also delete and redraw. Further, the annotation view also has a new "False Positive" button that lets the user indicate that a detected synapse is an FP. ##### Checkout If the user annotated all instances or marked them as FP he can click on checkout. At checkout, he sees a loading screen that indicates our model's process that calculates the 3D annotations based on the 2D seeds. Additionally, the user can download his 2D annotations as a backup and for manual use with our model. ##### Finish View indicating that all processes are done. Lets the user download the result and return to the landing page.
Overdue by 3 year(s)•Due by June 12, 2022•5/5 issues closedExtend the app with a view that lets the user categorize the defective segmentations.
Overdue by 3 year(s)•Due by May 10, 2022•2/2 issues closed# Improving SynAnno's overall performance Room for improvement: - Calculation of the bounding boxes - Calculation of the rotation angels - Loading of the data to the front end - Background transparency for the target labels
Overdue by 3 year(s)•Due by May 25, 2022•3/4 issues closed