Perfect Half Million Beauty Product Image Recognition Challenge at ACM Multimedia 2018 Link
Repo contains the scripts used for downloading the dataset and evaluation of our submission for the challenge [Team: VCA]
.
├── caffe-prototxt # Prototxt file for caffe model using inception-v1
├── dataset # test images urls and labels
├── scripts # scripts for evaluation on testset
├── utils # Helper code for retrieval and testing scripts
└── README.md
- Download the whole retrieval dataset consisting of 500K images in dataset folder
$ python crawl.py 0
-
Download Validation Set, Validation Ground Truth and Testset
-
Modify testbed.cfg file with location of dataset and caffe prototxt and run the following command to evaluate (Model weights are not shared)
$ python -u testbed.py --p deploy.caffemodel -i 0 | tee log.txt