serve icon indicating copy to clipboard operation
serve copied to clipboard

Torchserve unable to handle cv2 images

Open RutujaWanjari opened this issue 3 years ago • 1 comments

🐛 Describe the bug

I did some image composition ie overlaying 1 image on top of other using cv2

After running this command I get different results on local vs torchserve - cv2.imwrite("composited_image.png", composited_image)

Local machine result - composited_image_0

After running same code in torchserve - composited_image_0

Error logs

There are no error logs

Installation instructions

pip install torchserve pip install torch-model-archiver

Model Packaing

MAR packaging works totally fine

config.properties

No response

Versions

pip freeze -

aiohttp==3.8.1
aiosignal==1.2.0
anyio==3.6.1
asgiref==3.5.2
async-timeout==4.0.2
asyncer==0.0.1
attrs==21.4.0
beautifulsoup4==4.11.1
black==22.6.0
boto3==1.24.33
botocore==1.27.33
brotlipy==0.7.0
certifi @ file:///opt/conda/conda-bld/certifi_1655968806487/work/certifi
cffi @ file:///opt/conda/conda-bld/cffi_1642701102775/work
charset-normalizer==2.0.12
click==8.0.3
cryptography @ file:///tmp/build/80754af9/cryptography_1652101588893/work
cycler==0.11.0
docopt==0.6.2
enum-compat==0.0.3
fastapi==0.72.0
filelock==3.7.1
filetype==1.0.9
flatbuffers==2.0
fonttools==4.34.2
frozenlist==1.3.0
future==0.18.2
gdown==4.4.0
h11==0.13.0
idna @ file:///tmp/build/80754af9/idna_1637925883363/work
imageio==2.19.3
imutils==0.5.4
jmespath==1.0.1
kiwisolver==1.4.3
llvmlite==0.38.1
marshmallow==3.15.0
matplotlib==3.5.2
mkl-fft==1.3.1
mkl-random @ file:///tmp/build/80754af9/mkl_random_1626186066731/work
mkl-service==2.4.0
multidict==6.0.2
mypy-extensions==0.4.3
networkx==2.8.4
ninja==1.10.2.3
numba==0.55.2
numpy @ file:///opt/conda/conda-bld/numpy_and_numpy_base_1652801679809/work
onnxruntime==1.10.0
opencv-python==4.6.0.66
packaging==21.3
pathspec==0.9.0
Pillow==9.0.1
pipreqs==0.4.11
platformdirs==2.5.2
protobuf==4.21.2
psutil==5.9.1
pycparser @ file:///tmp/build/80754af9/pycparser_1636541352034/work
pydantic==1.9.1
PyMatting==1.1.5
pyOpenSSL @ file:///opt/conda/conda-bld/pyopenssl_1643788558760/work
pyparsing==3.0.9
PySocks @ file:///tmp/build/80754af9/pysocks_1605305812635/work
python-dateutil==2.8.2
python-multipart==0.0.5
PyWavelets==1.3.0
rembg==2.0.21
requests @ file:///opt/conda/conda-bld/requests_1657734628632/work
s3transfer==0.6.0
scikit-image==0.19.1
scipy==1.8.0
six @ file:///tmp/build/80754af9/six_1644875935023/work
sniffio==1.2.0
soupsieve==2.3.2.post1
starlette==0.17.1
tifffile==2022.5.4
tomli==2.0.1
torch==1.12.0
torch-model-archiver==0.6.0
torchaudio==0.12.0
torchserve==0.5.1
torchvision==0.13.0
tqdm==4.62.3
typing_extensions==4.3.0
urllib3==1.26.10
uvicorn==0.17.0
watchdog==2.1.7
yarg==0.1.9
yarl==1.7.2
zipp==3.7.0

Repro instructions

bash create_mar.sh
torchserve --start --ts-config torchserve.properties

Possible Solution

No response

RutujaWanjari avatar Jul 27 '22 11:07 RutujaWanjari

@RutujaWanjari Can you please share the steps to repro the problem. How are you passing the image, what model you are using, the code you have written etc. Thanks

agunapal avatar Jul 27 '22 16:07 agunapal

Please reopen the issue when you get a chance

agunapal avatar Aug 22 '22 17:08 agunapal