Team Leader Email - amitjoshi2052003@gmail.com
This section must include UML Daigrms and prototype description
List Down all technologies used to Build the prototype Clearly mentioning Intel® AI Analytics Toolkits, it's libraries and the SYCL/DCP++ Libraries used
- We take two datasets from cityscape website. One containing images and another containing json file which contain the bounding boxes for the objects in the image.
- We relate the files in both dataset and extract the boudning boxes from the json along with path for the image containing the objects.
- We store this data in a csv to be loaded again.
- We load the store csv file into a dataframe.
- We used yolo5 for this dataset.
- We processed the data for yolo by extracting the stored information, normalizaing the bounding boxes associated values between 0-1 and then storing them in respective directory structures required for yolo.
- We then pass the hyperparameters and config required to train our custom data.
- We ran the model for 100 epochs.
- We then can run the model of separate data inputs.
Write about the biggest learning you had while developing the prototype