Goal: The products were categorized into 3 groups: pharmacy, home and beauty, and the goal of the project is to detect and classify goods in a given image.
Implementation: Developed input-pipeline to feed the input data. Designed and trained a model using Tensorflow object detection api and SSD-Incpetion pre-weights Achieved mAP of 0.35.
Following are some of the inference images: For the Input image
Predictions:
more examples:
