cnclip

update readme

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