# README
Tiny YOLO v2
This is an example of Tiny YOLO v2 neural network. You can read about this network here.
Folder model contains yolov2-tiny.cfg on which file feedforward.go based.
Folder data contains image file dog_416x416.jpg - this is scaled to 416x416 image for make it better understanding of how net works.
For now this example works not quite fast (it is only "tiny" version). Probably it is better to implement this network using CUDA.
How to run:
go run .
# Functions
DenormalizeWeights Denormilize biases and kernels.
Float32bytes float32 to []byte.
Float32frombytes []byte to float32.
GetFloat32Image Returns []float32 representation of image file.
Image2Float32 Returns []float32 representation of image.Image.
IOUFloat32 Intersection Over Union.
MaxFloat32 Finds maximum in slice of float32's.
MaxInt Maximum between two integers.
MinInt Minimum between two integers.
NewTinyYOLOv2Net Constructor for TinyYOLOv2Net.
ParseTinyYOLOv2 Parse darknet weights (v2).
PrepareBiases Biases preparations.
PrepareConv Prepare convolutional kernels.
PrepareData Prepare kernels, biaseses, means, gammas and vars.
Rectify Creates rectangle.
Sigmoid Implementation of sigmoid function.
Softmax Implementation of softmax.
# Structs
DetectionRectangle Representation of detection.
TinyYOLOv2Net Tiniy YOLO v2 architecture.