基于深度神经网络的图像分类任务大作业.zip

# DeepIC Course Project for Machine Learning, Tsinghua University ### Task 12-class Image Classification using Deep Learning Technique. ### Approaches **Model: **Enhanced Convolution Neural Network We will try to use the following technique to enhance our model. 1. SpatialPyramidPooling: Used for enabling the network to accept images with arbitrary size. 2. SaliencyELD: Used for identifying the major semantic object within a image. 3. InceptionBlock: Inspired from ResNet, we plan to use this to ease the training process. 4. SIFT: Used as a supplement for convolution features. 5. Ensemble will also be adopted to improve overall model performance. ### Notice Place the dataset into ```data``` directory, it will not be uploaded. ### LICENSE [MIT](https://github.com/HeavenDuke/DeepIC/blob/master/LICENSE)