Weekly Reports

Final Report

Xinyu Zhao: This week I worked on the webapp with the group, doing webpage and testing of the data transfer. I also finished the project poster.   Tianhan Hu: This week I finished deploying webapp using flask and aiohttp on a linux server. The webapp version contains the basic functionality Read more…

Tianhan: This week I did a lot of benchmark on my trained hand detection model. I also trained another model using faster rcnn as backbone to compare the outcome. It turns out that the model trained with faster rcnn was actually too slow for our purpose (less than 1 fps), Read more…

Tianhan: This week I finished training our hand-detection model. I did some experiment and validation on it, and it seems pretty OK. Currently I am writing code to merge the detection model with the OpenCV part. I also looked more into how to launch a python webapp. It seems that Read more…

Xinyu Zhao: This week I kept finalizing working with contours and after modifying the epsilon when using the approximate function, and testing extensively both in static images and camera captures, it correctly outputs the coordinates of the shape, and it seems this is also more accurate than Hough line transform. Read more…

Tianhan: This week I continued to follow the online tutorial on transfer training with COCO models. I have fixed my training method using Google Cloud Platform. Hopefully I could finish the tutorial early next week, and train out our hand-detection model by the end of next week. Xinyu Zhao: This Read more…

    Team Update: Risk is that the coordinates capture by the first step might have some inaccuracies, and we have to also make sure that the data are correctly transferred to the AR part. We plan to have extensive trials to improve accuracy. No major change in design or Read more…

Tianhan: This week I continued to work on machine learning based method for hand detection. After comparison of several different methods, I have settled on using Tensorflow’s Object Detection API and the retraining method to solve the problem.As for the dataset, I decided to use the egohand dataset since it is Read more…

Team status update: This week’s progress was shorter because of spring break, and it’s slightly behind schedule   Tianhan Hu: This week I kept on researching on machine learning based approach for hand-detection. I was a little bit behind schedule due to midterm last week and spring break. Will try Read more…