SatarkAI/requirements.txt at main · bishowdip/SatarkAI · GitHub

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
# SatarkAI — Python dependencies (CPU-only, offline-deployable, Python 3.10/3.11)
# Install CPU PyTorch FIRST and explicitly, so pip does not pull the CUDA build:
#   pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu
# Then:  pip install -r requirements.txt
# See docs/01_ENVIRONMENT_AND_PREP.md

# --- API / backbone (required) ---
fastapi
uvicorn
python-multipart

# --- ML runtime (install torch/torchvision from the CPU index, see note above) ---
torch
torchvision
transformers
efficientnet_pytorch
timm

# --- Vision / image IO ---
pillow
numpy
opencv-python
facenet-pytorch          # MTCNN face detection (NOT dlib)
grad-cam                 # heatmap explainability (in-house Grad-CAM is fallback)

# --- Hashing / reuse detection ---
imagehash                # pHash perceptual hashing

# --- Metadata forensics ---
piexif
exifread

# --- Audio / video decode (ffmpeg/ffprobe is a separate bundleable binary) ---
librosa
soundfile

# --- Bilingual report ---
jinja2
weasyprint            # OPTIONAL: nicer HTML->PDF; needs system libs (pango/cairo)
reportlab             # offline PDF fallback — bundles Noto Sans Devanagari (reports/fonts/),
#                       so Nepali renders WITHOUT WeasyPrint/system libs (Windows-friendly)

# --- Content provenance (OPTIONAL) ---
c2pa-python           # verifies signed C2PA / Content Credentials manifests; without it,
#                       metadata.py falls back to an unverified byte-scan
# --- Translation (EN->NE report/UI; online + on-disk cache; optional) ---
deep-translator