[docker] Add docker with demo file#8
Conversation
Add a dockerfile which installs the prereqs. When built and ran, User will automatically enter a tmux session with FAN predicting the Astronaut sklearn image When built and ran, User will automatically enter a tmux session with FAN predicting the Astronaut sklearn image, this behaviour can be changed in future commits. Detector used for now is simply the HaarCascade from opencv. Next steps: - Either build a flask server with UploadImage button, detects the face and draws the landmarks - Or a jupyter notebook with a similar functionality - Script the model, currently I cannot do it because _modules is an issue in HourGlass - I haven't yet tested on GPUs, I believe I need to detect the gpu in bash and pass them using --gpus all
|
We tried to make the repo OS agnostic, but it seems the dockerisation will only work on Linux, right? Of course that's fine, but probably worth mentioning in the README. Also, I think it will be nice to use ibug.face_detection as the detector, as we tried put all our face analysis libs into the same namespace. |
|
OS: So far I've only tested on MacOS, but it should be the same as Docker can run on [Mac, Linux, Windows]. I'll test on my Windows machine soon and figure out if it's seamless too. |
Still not able to run on GPU but it's not a huge issue as demoing seems fine on a CPU, it takes .3 secs per image. Might not work for all versions of windows or environments, but I guess it's a working progress. Next steps: - Add the jupyter notebook demo - Figure out how to pull RetinaFace.
…t into dockerization
|
|
|
[x] Verify on Windows |
2 participants
Add a dockerfile which installs the prereqs.
When built and ran, User will automatically enter a tmux session with FAN predicting the Astronaut sklearn image When built and ran, User will automatically enter a tmux session with FAN predicting the Astronaut sklearn image, this behaviour can be changed in future commits.
Detector used for now is simply the HaarCascade from opencv.
Next steps: