Skip to content

Releases: michhar/yolov4-darknet-notes

Initial release

17 Sep 00:10
51bbca9
Compare
Choose a tag to compare
Initial release Pre-release
Pre-release

This is the initial release of a set of informal notes on running Darknet on an Azure VM to train a custom YOLO v4 model. Attached in Assets is a sample float16 quantized .tflite model (converted from a tiny YOLO v4 Darknet model), just for testing as it is not very accurate and only for 1 class: golden snapper (L. johnii). This model is not for production use and is only for demo usage.

Also, attached to this release in the Assets is the sample video to use to create the RTSP server for simulating an RTSP stream. This video will be used in a docker container. It has the audio removed and is in .mkv format, one of the supported formats for this server.

In addition, the anchor boxes will be needed for the final AI container and are:

[ 81, 27, 28, 80, 58, 51, 76,100, 109, 83, 95,246]