Contains previous projects with *.data, *.names, and *.cfg files
Contains python code to collect images using Bing API Contains python script to grab list of labels created on Labelbox Contains python script to download image from labelbox and transform label information into yolo format Contains python script to split labeled data to 80/10/10
Put your keys here if you need to access API with keys Bing requires API key to use their search labelbox API requires key to extract label information as json stream (alternative: download from website)
Contains python script to evaluate trained model
- git clone repository to under darknet folder where darknet was cloned and compiled
- configure dataprep.properties to accordingly (details tbd)
- Run dl_bingAPI_images.py to grab images using Azure cognitive services for images
- Run dl_labels_tocsv.py to grab label info from Labelbox.com
- Run prep_labeled_data.py to prepare labeled data and images for yolo format
- Run split_stage_data.py to split data into train set, valid set, and test set (80/10/10)
- To train configure and Setup your yolov3.cfg, obj.names, and obj.data, and download darknet53.conv.74 weights
- Run in command line interface: ./darknet detector train name_of_your_obj.data name_of_your_yolov3.cfg darknet53.conv.74 -dont_show -map a. -dont_show hides detailed errors from output b. -map uses mean average precision to train and uses valid data set c. -clear to train from the beginning
- After Training, to perform some spot checking use the following command: a. ./darknet detector test your.data your.cfg backup/your_final.weights data_for_test/image_name.jpg
- Configure model_eval.properties and run eval_mode_iou.py to view model performance