can't get segnet-camera to work

[gillamkid]

Using the guide on https://github.com/dusty-nv/jetson-inference/blob/master/docs/segnet-console.md I was able to get segnet-console working on my personal machine using the pre-trained aerial model. Here is the exact command I used (slightly different than the command in the tutorial):

./segnet-console drone_0255.png sam_output.jpg \
--prototxt=networks/FCN-Alexnet-Aerial-FPV-720p/fcn_alexnet.deploy.prototxt \
--model=networks/FCN-Alexnet-Aerial-FPV-720p/snapshot_iter_10280.caffemodel \
--labels=networks/FCN-Alexnet-Aerial-FPV-720p/fpv-labels.txt \
--colors=networks/FCN-Alexnet-Aerial-FPV-720p/fpv-deploy-colors.txt \
--input_blob=data \
--output_blob=score_fr_21classes

The output image comes out as expected with the overlay differentiating the sky from the ground.

I slightly modified the first line of the command as follows to be a segnet-camera version of the command:

./segnet-camera \
--prototxt=networks/FCN-Alexnet-Aerial-FPV-720p/fcn_alexnet.deploy.prototxt \
--model=networks/FCN-Alexnet-Aerial-FPV-720p/snapshot_iter_10280.caffemodel \
--labels=networks/FCN-Alexnet-Aerial-FPV-720p/fpv-labels.txt \
--colors=networks/FCN-Alexnet-Aerial-FPV-720p/fpv-deploy-colors.txt \
--input_blob=data \
--output_blob=score_fr_21classes

No errors are printed to the console, and a popup appears where I can see the live video from the TX2 dev board camera, however there is no overlay at all. No status bar, no change in coloration.

I've tried using a webcam. Didn't work. I Googled everything I could think of but have not found any posts from someone who was able to get segnet-camera working.

I'm using a Jetson TX2 that I just imaged with the most recent version of Jetpack yesterday.

[gillamkid]

I tried it on a different TX2, one with an older version of Jetpack. I tried running it normally and then with overclocking. Still no overlay.

[gillamkid]

Solved the problem. segnet-camera.cpp displays imgRGBA (the original video image) and not outCUDA (the overlayed video image). I'll make a pull request.

[gillamkid]

Here's the pull request
#386