Summary
Methods to convert TensorFlow* models that are trained with images in RGB* order
Description
- Correct result when inferencing with original Keras model.
- Converted the Keras model into Intermediate Representation.
- Wrong result when inferencing with the converted model.
Resolution
Most TensorFlow models are trained with images in RGB order.
By default, Inference Engine samples and demos expect input with BGR channels order.
Method 1:
Convert the model using the Model Optimizer tool with --reverse_input_channels argument.
Method 2:
Change from RGB to BGR:
im_cv = cv.imread('test2.jpg')
frame = cv.cvtColor(im_cv, cv.COLOR_RGB2BGR)
Additional information
Refer to When to Reverse Input Channels