Sprint goal setting

RESEARCH GOALS

1. Survey on uncertainty estimation methods for Keypoint detection
2. Comparison of uncertainty estimation methods for Keypoint detection
3. Implementation of uncertainty estimation methods on robot grasping application

---

SPRINT GOALS

---

October 15

1. Development of a Gaussian uncertainty model and implementing it on the Kinovo robot.

   • Without albumentations
   • Finding best pose for grasping using uncertainty metric
   • Plot uncertainty distributions on images
   • Try introducing minor albumentation techniques in the dataloader and check the performance
   • Robot implementation

2. Comparison between Gaussian and Generalised Gaussian

3. Mid term report (UCI, Comparison)

---

September 15

1. ✓ Submit thesis application
2. Train model for two loss functions
3. Plot evaluation results for Cornell Grasp Dataset
4. Report on UCI datasets results in a paper format
5. ✓ Add neptune logging
6. Code review of training pipeline
7. Revise research goal

---

August 15

1. ✓ Work on UCI datasets
2. ✓ Plot comparison Graphs - Interval score & RMSE
3. ✓ Dataloader for Cornell Grasp Dataset
4. ✓ Proposal completion

---