Library for simple object recognition on images using YOLO model.

Inspired by machinelearningmastery.com article by Jason Brownlee.

• as a library:

from recogn_img import read_classes, Recognizer

classes = read_classes("coco_classes.txt")
recognizer = Recognizer(model_path="yolov3.h5",  classes=classes,
                        img_width_height=(416, 416),
                        threshold_box_obj=(0.5, 0.6))
results = recognizer.recognize("my_image.jpg")
print(results)

• as CLI tool (single JSON with results):

recogn-img yolov3.h5 coco_classes.txt ~/Pictures ~/RecognitionResults.json --img-width 416 --img-height 416 --box-threshold 0.5 --obj-threshold 0.6

• as CLI tool for storing images with rendered boxes in given dir:

render-recogn ~/RecognitionResults.json ~/RenderedImgs/ --copy-exif

• Linux with Python >= 3.7 installed
• CPU with AVX instruction extensions (tensorflow>=1.5.1 requires them)
• YOLO pre-trained model (weights file) converted to Keras model (.h5 file), and file with list of classes for given model, both downloaded through make dl_model command for your convenience
• system libs (ubuntu example: sudo apt update && sudo apt install -y libsm6 libxext6 libxrender1)

• for development, take a look at Makefile