Tiny YOLOv2

**Paper** [YOLO9000: Better, Faster, Stronger][1] **Dataset** The Tiny YOLO model was trained on the Pascal VOC [dataset.][2] **Source** The model was converted from a Core ML version of Tiny YOLO using [ONNXMLTools][4]. The source code can be found [here][5]. The Core ML model in turn was converted from the [original network][6] implemented in Darknet (via intermediate conversion through Keras). **Model input and output** **Input** * Input image of the shape (3x416x416) **Output** * Output is a (1x125x13x13) array **Pre-processing steps** Resize the input image to a (3x416x416) array of type float32. **Post-processing steps** The output is a (125x13x13) tensor where 13x13 is the number of grid cells that the image gets divided into. Each grid cell corresponds to 125 channels, made up of the 5 bounding boxes predicted by the grid cell and the 25 data elements that describe each bounding box (5x25=125). For more information on how to derive the final bounding boxes and their corresponding confidence scores, refer to this [post][7]. **Sample test data** Sets of sample input and output files are provided in .npz format (test_data_*.npz). The input is a (3x416x416) numpy array of a test image from Pascal VOC, while the output is a numpy array of shape (1x125x13x13). **Model size: 5.3 MB** **License** [MIT][8] [1]: https://arxiv.org/pdf/1612.08242.pdf [2]: http://host.robots.ox.ac.uk/pascal/VOC/ [3]: https://pjreddie.com/darknet/yolov2/ [4]: https://github.com/onnx/onnxmltools [5]: https://github.com/hollance/YOLO-CoreML-MPSNNGraph [6]: https://pjreddie.com/darknet/yolov2/ [7]: http://machinethink.net/blog/object-detection-with-yolo/ [8]:https://github.com/onnx/models/blob/master/LICENSE