Integration of deepfake detection models#29
Integration of deepfake detection models#29lous-e wants to merge 53 commits intoUMass-Rescue:mainfrom
Conversation
added model parameter options, updated the server, etc.
* deepfake pyproject.toml * improved sim_data.py 's dataset_path (line 20) * renamed server.py to main.py * Fixed model paths in each of the model process files * updated imports to be deepfake_detection.NAME * fixed models loading before specification of which models * fixed add_ml_service * fixed schema functions
Some fixes
| ollama = ">=0.4.7,<0.5.0" | ||
| pypdf2 = ">=3.0.1,<4.0.0" | ||
| pandas = ">=2.2.3,<3.0.0" | ||
| pillow = ">=11.2.1,<12.0.0" |
There was a problem hiding this comment.
notice that this Python package has a special license open source MIT-CMU License req
https://github.com/python-pillow/Pillow/blob/main/LICENSE
please add this too
src/deepfake-detection/README.md
Outdated
| ### Download and run RescueBox Desktop from the following link: [Rescue Box Desktop](https://github.com/UMass-Rescue/RescueBox-Desktop/releases) | ||
|
|
||
| #### Open the RescueBox Desktop application and register the model | ||
|  |
There was a problem hiding this comment.
new UI does not have a register option please remove this
also remove old UI images src/deepfake-detection/deepfake_detection/images/*.png
|
|
||
| [tool.poetry.dependencies] | ||
| numpy = "*" | ||
| pandas = "*" |
There was a problem hiding this comment.
The pandas package is licensed under the BSD 3-Clause "New" or "Revised" License. This license permits the use, ...provided that the copyright notice, conditions, and disclaimer are retained.
please add this license file
| readme = "README.md" | ||
|
|
||
| [tool.poetry.dependencies] | ||
| numpy = "*" |
There was a problem hiding this comment.
https://numpy.org/doc/stable/license.html
please add this license
|
|
||
|
|
||
| # Personal Attributes (Stuff about the subject) | ||
| possible_genders = ["Male", "Female"] # Consider wording Man/Woman or Boy/Girl |
There was a problem hiding this comment.
did you consider "unknown" as an option ?
(for all these prompts ? )
There was a problem hiding this comment.
Yes. The intent of each attribute is to diversify the results. "Unknown" or "None" was too non descriptive in my eyes since the model would just assign the most obvious one(s). So I elected to omit that type of attribute category.
| name="Image DeepFake Detector", | ||
| author="UMass Rescue", | ||
| version="0.2.0", | ||
| info=load_file_as_string("img-app-info.md"), |
There was a problem hiding this comment.
the pull path to img-app-info.md has to be passed to load_file_as_string() method
example src/face-detection-recognition/face_detection_recognition/face_match_server.py
BUG: message_when_empty="Select the models to use for prediction, default is all.", message when empty in rescuebox desktop is just empty--no message at all?
…us-e/RescueBox into facecropping-and-last-changes
Facecropping and last changes
lous-e
changed the title
6 participants
Libraries (versions represent those added in the original requirements.txt file
numpy
pandas
pillow
onnxruntime
Models:
ViT
ViT
BNext
BNext