Goh Jia Yi, Jesa

@yosagaf Hi, thank you for your help! In legday.py I changed these few lines `N=1`, `feature_vec = np.zeros(36)`, `ff = np.expand_dims(np.concatenate(list(ffwd_n_avg(q)), axis=0).reshape(36, N), axis=0)`. Basically changing the window and pose...

@yosagaf Nice I managed to solve the keras version mismatch. I'll try out the repo that you have linked, thank you so much for your help.

Hope this isn't too late! First, collect video data for the actions you want to detect. Second, you'll have to edit the `iva.py` script. There is a section of code...

Just tried it today and doesn't work, well can't do much when FB's UI is constantly changing..

在 `setup.py` 里发现需要用 TF 1.4.0-1.9.0, 就只可以用CUDA 9。想用CUDA 10就必须升级Tensorflow,查看这[链接](https://www.tensorflow.org/install/source#tested_build_configurations)。 Within `setup.py` the requirements for TensorFlow is 1.4.0 to 1.9.0. According to the build guide [here](https://www.tensorflow.org/install/source#tested_build_configurations) we can only use CUDA 9....

For me, I was unable to get a video feed when I run the cell that was supposed to create a widget used to display the camera feed with visualizations...

Hi, it took me ~5 mins to run on a Jetson Nano. I ran the exact same code on `trt_pose/tasks/human_pose/live_demo.ipynb`. Looks like your Nano ran out of memory, perhaps consider...

> According to the note, if you delete following lines in selfModules/neg.py, it will work just fine. > assert parameters_allocation_check(self), > """ > Invalid CUDA options. out_embed and in_embed parameters...

I agree that NIMA is rather practical, art is subjective even for humans. Interesting insights you have gathered!! One possible high-level explanation I have for the difference in results -...

One more thing is to check your CUDA and cuDNN versions to make sure that they are compatible [here](https://www.tensorflow.org/install/source#tested_build_configurations). Here's how you can [check your CUDA version](https://stackoverflow.com/questions/9727688/how-to-get-the-cuda-version/9730706#9730706) and [change it...