I am using "mobilenetv2 ssd fpnlite 320x320" to train my own object detection model build from scratch. However, I am trying to use grayscale images for both training and deploying the tflite model. I know that even when I input 1-channel images, the output becomes 3-channel by duplicating one channel into three. I find this unnecessary and want to avoid wasting memory and CPU usage, especially since I plan to deploy my model in grayscale. Which file must I change to create a 1-channel model instead of a 3-channel model?
3. Steps to reproduce
Inside of ssd_mobilenet_v2_fpn_keras_feature_extractor.py, I change
4. Expected behavior
I expecting shape: [1 320 320 1] but it print [1 320 320 3]
Prerequisites
Please answer the following questions for yourself before submitting an issue.
1. The entire URL of the file you are using
https://github.com/tensorflow/models/tree/master/official/...
2. Describe the bug
I am using "mobilenetv2 ssd fpnlite 320x320" to train my own object detection model build from scratch. However, I am trying to use grayscale images for both training and deploying the tflite model. I know that even when I input 1-channel images, the output becomes 3-channel by duplicating one channel into three. I find this unnecessary and want to avoid wasting memory and CPU usage, especially since I plan to deploy my model in grayscale. Which file must I change to create a 1-channel model instead of a 3-channel model?
3. Steps to reproduce
Inside of ssd_mobilenet_v2_fpn_keras_feature_extractor.py, I change
4. Expected behavior
I expecting shape: [1 320 320 1] but it print [1 320 320 3]
5. Additional context
My config look like....
6. System information