Ethan Muchnik Status Report 11/09

• Most of my work this week related to optimizing the statistics generating script for each model configuration and fixing minor bugs. Essentially I found minor bugs in said script and some stats that were lost after training was completed. So I completely refactored the Fiftyone.py script to Train.py, Validation.py and PreprocessData.py.
• I also began the time-consuming work of going through existing COCO photos and getting the ones that had unlabelled pencils/pens in them. To make the work less time-consuming I got the smallest possible size of COCO model that still had an mAP of above 50%.
• I also tested the new model on the Rasp Pi more rigorously in various locations that were not in the training data and got good results. I also verified that the rasp pi could be set up and the code could be run in various rooms in Hamerschlag.
• Due to the huge refactoring efforts, GroundingDino testing/development was put on hold for a good chunk of this week.

Is your progress on schedule or behind? If you are behind, what actions will be
taken to catch up to the project schedule?

My progress for GroundingDino exploration is a bit behind due to other things taking up my time this week though I now believe I’m in a good place to develop in that realm. At some point in the next 1.5 weeks I’ll just have to accept the model that I have at the time and work on the other components.

What deliverables do you hope to complete in the next week?

I am aiming to release graphics comparisons of the YOLO vs Grounded DINO and then perhaps hybrid approaches. This was initially planned for last week though the bugs in the stats/validation scripts caused a delay.

I would like to complete further testing when photos with unlabelled pencils/pens/wallets are labelled.

If time permits I would like to work on the audio to text processing.