Releases: ZHEQIUSHUI/CLIP-ONNX-AX650-CPP
Releases · ZHEQIUSHUI/CLIP-ONNX-AX650-CPP
cnclip
3models
将image encode和text encode解耦出来更灵活高效
./main --ienc image_encoder.onnx --tenc text_encoder.onnx --dec feature_matmul.onnx -v ../vocab.txt -i ../images/ -t ../text.txt
inputs:
images: 1 x 3 x 224 x 224
output:
image_features: 1 x 512
decode Inference Cost time : 0.00040005s
per image:
image path\text| bird| cat| dog|
../images/bird.jpg| 1.00| 0.00| 0.00|
../images/cat.jpg| 0.00| 0.99| 0.01|
../images/dog-chai.jpeg| 0.00| 0.00| 1.00|
per text:
text\image path| ../images/bird.jpg| ../images/cat.jpg| ../images/dog-chai.jpeg|
bird| 0.87| 0.01| 0.12|
cat| 0.00| 0.98| 0.02|
dog| 0.00| 0.00| 1.00|
./main --ienc image_encoder.axmodel --tenc text_encoder.onnx -d feature_matmul.onnx -v vocab.txt -t text.txt -i images/
Engine creating handle is done.
Engine creating context is done.
Engine get io info is done.
Engine alloc io is done.
[I][ init][ 275]: RGB MODEL
decode Inference Cost time : 0.000754583s
per image:
image path\text| bird| cat| dog|
images/bird.jpg| 1.00| 0.00| 0.00|
images/cat.jpg| 0.01| 0.98| 0.01|
images/dog-chai.jpeg| 0.00| 0.00| 1.00|
per text:
text\image path| images/bird.jpg| images/cat.jpg| images/dog-chai.jpeg|
bird| 1.00| 0.00| 0.00|
cat| 0.00| 0.99| 0.01|
dog| 0.00| 0.00| 1.00|
models
add ax650 backend