Object detection with tensorflow API

configuring object detector Training Pipeline

trainig pipiline can be found in object_detection/protos/pipeline.proto

1.Picking Model parameters

sample models configurations : object_detection/samples/configs folder

note: num_classes should be changed to a suitable value

2.Defining input

input format:TFRecord file format
specify a label map: mapping between a class id and class name

3.Config the Trainer
Model parameter initialization
Input preprocessing
SGD parameters

4.Config the Evaluator
metrics as defined by PASCAL VOC challenge