Improvements to StereoDNN?
v-raja opened this issue · comments
Will improvements to StereoDNN be made to improve performance on the TX2 anytime soon?
Also, I was wondering if it would be possible to implement YOLO v3 in the StereoDNN ROS node? If so, could you please give me tips on how I could update the StereoDNN node?
Thanks!
We currently do not have any improvements planned for StereoDNN.
If you are interested in Jetson-tested object detection models, this repository should hopefully help. The repo has SSD model which, for any practical scenario, should work as well as YOLO.
@Alexey-Kamenev If I'm using a ZED camera, do I need to re-train the model or can I just use your models from the repo? And what resolution of images from the ZED camera gives the best depth accuracy? (Sorry for the novice questions!)