-
Notifications
You must be signed in to change notification settings - Fork 0
/
notes.todo
72 lines (62 loc) · 3.35 KB
/
notes.todo
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
DEVELOPEMENT
------------
- [x] Fork only opencv or opencv_contrib to make the changes the other repo fetch the upstream
- [x] Containerized builds. Build one container which builds and installs Opecv with Python bindings. The build another container inheriting the first and installing the app. So the firsta container will not change all the time.
Containers built:
- [x] First container build and install opencv. Problem with luck of space and not finding numpy
- [x] Second container build and install the django application
- [X] Fix the HTTP API according to the document
- [x] Fix video codes on OpenCV to read all kind of videos
- [x] Create aggragated metadata from every annotation process in the form of shotlist
- [x] Emebed Tensorflow for object detection instead of opencv_dnn module
- [x] Add authentication system -> Integration with SSO
- [x] Add docker volume to store data, and fix the whole set up
- [x] Option for bounding boxes (true/false)
- [x] Selective Search before object detection
- [x] Add celery for asynchronous tasks
- [x] Create tasks for face detection/ face recognition/ object detection
- [x] Fix celery tasks time limits to handle long detection tasks [1]
- [ ] API calls [0]
- [x] POST (upload) zip file with faces and create faces db and create pretained config
- [x] GET list of all models uploaded
- [x] GET default and example configuration
- [x] POST to forms and annotate
- [x] POST json to and annoatate
- [x] Better error messages if provided json is invalid!
- [ ] Train cascade + List Cascades
- [ ] Give the oppurtunity to train a haar cascade to recognize specific objects
- [x] Update LBPH recognizer trained data, instead of retrain
- [x] Integrate with MCSSR
- [x] Send back error messages
- [x] Fix SSO configuration
- [x] Add API tests on Jenkins
- [x] Train object recognizer with a bigger database [2]
- [ ] Add Amazon S3 support
- [ ] Process videos as bunch of frames in parallel
- [x] Automatically select every how many frames will run the annotation process
- [x] Give option to user to detect/recognize every X frames
- [x] Automatic subtitling (create subs with autosub tool, merge subtitles with ffmpeg)
- [x] Fix API call to accept input and output language
- [x] Send back annotation metadata in JSON to MCSSR
- [x] Create aggregated results either to DB or to another ext service after the finish of tasks
- [x] Create a two separate tasks Face detection/recognition + Object detection \
\ that can run autonomous and in parallel. They should have a parameter to read every X frames
- [x] Normalize face detection results
- [x] Normalize face recognition results
- [x] Check whether the CNN for object detection can recognize animals
-> Can recognize dog, cat, bird, horse, sheep, cow, elephant,
bear, zebra, giraffe
EVALUATION
----------
- [ ] Benchmarks
- [x] Screencast of usage
IMPROVEMENTS
------------
- [ ] Try to solve issues with py-faster rcnn
- [ ] Embedd py-faster rcnn to the application
- [ ] Gender classification
- [ ] Age classification
- [ ] Landmark/Location detection
[0] https://philsturgeon.uk/api/2016/01/04/http-rest-api-file-uploads/
[1] http://docs.celeryproject.org/en/latest/userguide/workers.html#time-limits
[2] https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md