Rocket Model Trainer

Instructions on how to create and train High Power Rocketry tracker model with Python and Yolo. The weights generated by this method can be used to run your custom tracker.

Todo

• Uploaded OID4-rocket with Yolov3 notations
• Added script to generate train.txt file,put the output file "train.txt" in "darknet-master\build\darknet\x64\data "folder.
• Create a "obj" folder inside the folder having "train.txt"
• Add all the images under train\images folder into "darknet-master\build\darknet\x64\data\obj" folder
• Add all the yolov3 labels txt files in train\label-yolov3 folder into "darknet-master\build\darknet\x64\data\obj" folder
• Create "obj.data" in "\data" folder with

classes = 1
train = data/train.txt
valid = data/test.txt
names = data/obj.names
backup = backup/

• Create "obj.names" with "0" inside
• Copy "yolov3-tiny_obj.cfg" from "x64\cfg" file to "x64". We will experiment with it later on.Rename it to "yolov3-tiny-obj.cfg". Modyfying it.

[net]
# Testing
#batch=64
#subdivisions=8

filters = number of classes5 + 3, in my case, 1 = rocket, 15+3

[convolutional]
size=1
stride=1
pad=1
filters=8
activation=linear

• [x]Also it is advised to calculate anchors of training dataset. Therefore, we will do that by opening terminal or cmd and running darknet_no_gpu.exe detector calc_anchors data/obj.data -num_of_clusters 6 -width 416 -height 416 hit enter and it should compute the anchors. Paste them inside the first yolo layer , like this

[yolo]
mask = 3,4,5
anchors = 9, 55,  23,127,  70,136,  48,268, 110,316, 279,304

Note: The -cluster is number of pairs, tiny yolo has 6. It can change

• Train custom yolov3 object detector
• run darknet_no_gpu.exe partial cfg/yolov3-tiny.cfg yolov3-tiny.weights yolov3-tiny.conv.15 15
• darknet_no_gpu.exe detector train data/obj.data yolov3-tiny-obj.cfg yolov3-tiny.conv.15 -dont_show

Todo - Build the project for windows

• Tried various methods. Easiest is to simply download opencv frol sourceforge. Extract into single folder with name "opencv" inside "opencv_3.0" folder. Build x64,release. Try darknet_no_gpu.exe detect cfg/yolov3.cfg yolov3.weights data/eagledarknet_no_gpu.exe detect cfg/yolov3.cfg yolov3.weights data/eagle

• Then open yolo-no-gpu with visual studio 15. Set: Release and x64 IS. Click second build tolo-no-gpu . Pray that it builds.

• finally something is working. Don't give up.