Sprint Goal Setting

Fifth Goal Set (Mid Term)-> 15/09/2022.

1. RQ1
   • Comparative evaluation of 2 uncertainty estimation techniques (evidential and ensembles) using uncertainy labels dataset of @work components.
2. RQ2
   • What are the modifiable parameters in blender
   • Which parameters of all the available options can be used/helpful for generating the datasets (classification and uncertainty labels)
3. Find code for ensembles uncertainty estimation technique and integrate it in your code.
4. Complete Texture baking process for @work components and save the files in .obj format and necessary images for applying textures later.
5. Modify the dataset generation code with requirements and refactor the code.

Fourth Goal set -> 15/08/2022.

1. ✔️ Generate synthetic dataset for two classes (50 or 100 images for each class) use same background with random rotation and random lighting. ----> select objects like M20 and M30.
2. ✔️ Create one simple uncertainty rule (lighting)
3. ✔️ Create code for generating uncertainty label for one simple uncertainty rule (lighting)
4. ✔️ Generate dataset for 2 classes along with uncertainty labels (50 or 100 images)
5. ✔️ Train DNN model using evidential loss and the 2 synthetic dataset classes along with uncertainty labels.
6. ✔️ Train DNN model using evidential loss and asus_combined dataset in cluster.
7. ✔️ Train DNN model using cross_entropy loss and 2 classes of synthetic dataset.
8. Answer research question for mid-term ---->
   • Which method can be used for determining the ground truth for uncertainty.
     • How do we generate ground truth values from the rule based criteria.
     • How do we map the uncertainites to the synthetic dataset.
     • What are the modifiable parameters in blender.

Third Goal Set -> 15/07/22

1. ✔️ Add scheduler in the DNN pipeline.
2. ✔️ Add augmentation techniques to the DNN pipeline.
3. ❌ Train the DNN model for evidential loss on the cluster for the ausus_combined dataset.
4. ✔️ Write a summary for the 3 papers and add questions.

Second Goal Set -> 15/06/2022

1. ✔️ Update Readme.md file
2. ✔️ Automate the complete pipeline for a Deep learning network including train, eval, and figures everything should be.py files and the main.py file convert from jupyter notebooks to python files. Don't use jupyter notebooks. Also add neptue logger(logging).
3. ✔️ Make a presentation on what is the problem and why is it important.
4. ✔️ Create something with blender using blender(bpy). Generating dataset. Learning bpy. Creating a list of parameters.
5. ❌ Create one simple fuzzy rule in python.
6. ❌ Understanding 3 uncertainty papers.

First Goal Set -> 11/05/2022

1. ✔️ Write the Draft Proposal
2. ✔️ Meeting with Prof.
3. ✔️ Something with blender
4. ❌ Understanding 5 uncertainty papers