Update readme.md

Author: yunlongdong

readme.md

@@ -1,24 +1,14 @@
-## infCNN
-A pure numpy-based inference framework for CNN. The infCNN supports the inference of model trained on pytorch.
+## numpy-cnn
+A pure numpy-based inference framework for CNN. The aim of numpy-cnn is to deploy CNN model with low-cost and few adjustment including embedded systems. Also in order to enlarge the application scope, we support ONNX format, which enables the converting of trained model within various DL frameworks (PyTorch).  
 
-In general, all the elements used in the inference of CNN are divided into ```op``` and ```layer```. ```op``` contains no trainable weights such as "relu, sigmoid, softmax, maxpool, flatten", ```layer``` contains trainable weights such as "conv2d, dense". The ```op``` and ```layer``` implemented are very few now.
+## Features
+We establish a extremely streamlined structure to build the CNN inference framework. All the elements (layers, operations, activation fuctions) are abstracted to be ```layer```, and a json formatted ```flow``` is applied to build the computation graph. And the weights are stored in ravel numpy, which largely simplifies the loading process. 
 
-#### Inference
-
-Example is shown in ```net.py```, which shows the inference of a LeNet CNN. Weights converted from pytorch model are loaded here. To be noted that ```net.py``` should align with ```train/lenet.py``` since they should have the same model. Just replacing ```nn.Conv2d/Linear``` with ```inferCNN.Conv2d/Dense```   will work.
-
-#### Training on pytorch
-
-The training on pytorch is regular, which can be seen in ```train/``` folder. The weights of CNN are exported to ```.mat``` file.
-
-
-#### Plugins for ImagePy
-The plugin example for mnist is in ```plgs/``` folder
-
-#### References
-
-* https://github.com/wiseodd/hipsternet
-* https://github.com/pytorch/examples/blob/master/mnist/main.py
+## Demos
+We have released some demos, which can be investigated inside ```demo/``` folder.
+![](https://raw.githubusercontent.com/Image-Py/numpy-cnn/master/demo/craft_text_detector/rst.png)
+![](https://raw.githubusercontent.com/Image-Py/numpy-cnn/master/demo/hed_edge_detector/rst.png)
+![](https://raw.githubusercontent.com/Image-Py/numpy-cnn/master/demo/resnet18/rst.png)