tensorflow-fcn-master

代码说明：

说明：  卷积网络正在推动着图像识别方面的进步，其不仅改进了整体图像的分类效果，而且在具有结构化输出的局部任务上也取得了进步，包括边界框目标检测，关键点预测等。 自然下一步是改进在像素级别上的预测。其实，以前的方法已经使用卷积网络进行语义分割任务，其中每个像素都被标记为属于目标或属于其他区域，但让具有缺点。 FCN和CNN的区别：CNN卷积层之后连接的是全连接层；FCN卷积层之后仍连接卷积层，输出的是与输入大小相同的特征图，提出一个端到端，像素对像素的全卷积网络用于语义分割任务(Convolution network is promoting the progress of image recognition. It not only improves the classification effect of the whole image, but also makes progress in the local tasks with structured output, including boundary box target detection, key point prediction and so on. The natural next step is to improve prediction at the pixel level. In fact, previous methods have used convolutional networks for semantic segmentation tasks, in which each pixel is marked as belonging to the target or other regions, but it has disadvantages. The difference between FCN and CNN: CNN convolution layer is connected with full connection layer after CNN convolution layer; FCN convolution layer is still connected with convolution layer after FCN convolution layer, and the output is the same as the input size of feature map. An end-to-end, pixel to pixel full convolution network is proposed for semantic segmentation task)

文件列表：

.gitignore, 782 , 2018-07-13
.idea, 0 , 2018-07-14
.idea\inspectionProfiles, 0 , 2018-07-13
.idea\misc.xml, 298 , 2018-07-13
.idea\modules.xml, 294 , 2018-07-13
.idea\tensorflow-fcn-master.iml, 441 , 2018-07-13
.idea\workspace.xml, 32561 , 2018-07-14
__pycache__, 0 , 2020-09-10
__pycache__\BatchDatsetReader.cpython-37.pyc, 3645 , 2020-09-10
__pycache__\read_MITSceneParsingData.cpython-37.pyc, 2090 , 2020-09-10
__pycache__\TensorflowUtils.cpython-37.pyc, 8757 , 2020-09-10
BatchDatsetReader.py, 4530 , 2018-07-13
Data_zoo, 0 , 2018-07-13
Data_zoo\MIT_SceneParsing, 0 , 2018-07-14
Data_zoo\MIT_SceneParsing\MITSceneParsing.pickle, 4705275 , 2018-07-13
FCN.py, 14464 , 2018-07-14
images, 0 , 2018-07-14
Model_zoo, 0 , 2020-09-10
read_MITSceneParsingData.py, 4188 , 2018-04-24
TensorflowUtils.py, 9735 , 2018-07-14