GitHub - minhnd3796/NGUYENDUCMINH

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 119 Commits
MatConvNet Tutorials		MatConvNet Tutorials
PapersAndSlides		PapersAndSlides
ResNet-101_Keras		ResNet-101_Keras
__pycache__		__pycache__
test_images		test_images
.FCN-4s-ResNet101-5c-v3.3.py.swp		.FCN-4s-ResNet101-5c-v3.3.py.swp
.FCN-8s-ResNet101-5c-v2.2.py.swp		.FCN-8s-ResNet101-5c-v2.2.py.swp
Batch_manager.py		Batch_manager.py
Batch_manager.pyc		Batch_manager.pyc
Batch_manager_5channels.py		Batch_manager_5channels.py
DenseBlock.jpg		DenseBlock.jpg
DenseNet.jpg		DenseNet.jpg
FCN-4s-ResNet101-5c-v3.3.py		FCN-4s-ResNet101-5c-v3.3.py
FCN-4s-ResNet101-6c-v3.3.py		FCN-4s-ResNet101-6c-v3.3.py
FCN-4s_Potsdam_submission.py		FCN-4s_Potsdam_submission.py
FCN-8s-ResNet101-5c-v3.0.py		FCN-8s-ResNet101-5c-v3.0.py
FCN-8s-ResNet101-5c-v3.1.py		FCN-8s-ResNet101-5c-v3.1.py
FCN-8s-ResNet101-5c-v3.2.py		FCN-8s-ResNet101-5c-v3.2.py
FCN8s_vgg19_potsdam.py		FCN8s_vgg19_potsdam.py
FCN_4s_ResNet101_5_v33.py		FCN_4s_ResNet101_5_v33.py
FCN_4s_ResNet101_6c_v33.py		FCN_4s_ResNet101_6c_v33.py
FCN_8s_ResNet101_5c_v32.py		FCN_8s_ResNet101_5c_v32.py
README.txt		README.txt
batch_eval_15.py		batch_eval_15.py
batch_eval_potsdam.py		batch_eval_potsdam.py
batch_eval_top.py		batch_eval_top.py
build_resnet_graph_15channels.py		build_resnet_graph_15channels.py
build_resnet_graph_potsdam.py		build_resnet_graph_potsdam.py
calculate_mean.py		calculate_mean.py
calculate_mean_potsdam.py		calculate_mean_potsdam.py
check_5_classes.py		check_5_classes.py
check_model.py		check_model.py
convert_feature_csv_to_image.py		convert_feature_csv_to_image.py
convert_gt2annotations.py		convert_gt2annotations.py
convert_gt2annotations_potsdam.py		convert_gt2annotations_potsdam.py
convert_to_tfrecord.py		convert_to_tfrecord.py
convert_to_tfrecord_potsdam.py		convert_to_tfrecord_potsdam.py
count_crop.py		count_crop.py
csv2image.py		csv2image.py
data_reader.py		data_reader.py
data_reader.pyc		data_reader.pyc
data_reader_5channels.py		data_reader_5channels.py
demo_FCN-ResNet101-4s.py		demo_FCN-ResNet101-4s.py
demo_FCN-ResNet101-4s_Potsdam.py		demo_FCN-ResNet101-4s_Potsdam.py
demo_FCN-ResNet101-8s.py		demo_FCN-ResNet101-8s.py
draft.m		draft.m
evaluate.py		evaluate.py
feat_ext.m		feat_ext.m
fully_conv_densenet.py		fully_conv_densenet.py
fully_conv_densenet_6channels.py		fully_conv_densenet_6channels.py
fully_conv_densenet_potsdam.py		fully_conv_densenet_potsdam.py
fully_conv_resnet.py		fully_conv_resnet.py
fully_conv_resnet.pyc		fully_conv_resnet.pyc
fully_conv_resnet_15_channels_3skips.py		fully_conv_resnet_15_channels_3skips.py
fully_conv_resnet_5_channels.py		fully_conv_resnet_5_channels.py
fully_conv_resnet_5_channels_v2.py		fully_conv_resnet_5_channels_v2.py
fully_conv_resnet_v2.py		fully_conv_resnet_v2.py
fully_conv_resnet_v3.py		fully_conv_resnet_v3.py
fully_convnets.py		fully_convnets.py
fully_convnets_15_channels.py		fully_convnets_15_channels.py
fully_convnets_5_channels.py		fully_convnets_5_channels.py
fully_convnets_5_channels_v2.py		fully_convnets_5_channels_v2.py
fully_convnets_5_channels_v3.py		fully_convnets_5_channels_v3.py
fully_convnets_potsdam.py		fully_convnets_potsdam.py
fully_convnets_v2.py		fully_convnets_v2.py
get_15_channel_images.py		get_15_channel_images.py
get_npy_Potsdam.py		get_npy_Potsdam.py
image2csv.py		image2csv.py
infer_ensemble.py		infer_ensemble.py
infer_ensemble_potsdam.py		infer_ensemble_potsdam.py
infer_ensemble_resnet.py		infer_ensemble_resnet.py
infer_ensemble_resnet_15.py		infer_ensemble_resnet_15.py
infer_ensemble_resnet_potsdam.py		infer_ensemble_resnet_potsdam.py
infer_imagenet_resnet_101.py		infer_imagenet_resnet_101.py
infer_imagenet_resnet_101.pyc		infer_imagenet_resnet_101.pyc
infer_imagenet_resnet_101_5chan.py		infer_imagenet_resnet_101_5chan.py
infer_imagenet_resnet_101_5chan_v2.py		infer_imagenet_resnet_101_5chan_v2.py
infer_imagenet_resnet_101_5chan_v3.py		infer_imagenet_resnet_101_5chan_v3.py
infer_imagenet_resnet_101_5chan_v31.py		infer_imagenet_resnet_101_5chan_v31.py
infer_imagenet_resnet_101_v2.py		infer_imagenet_resnet_101_v2.py
infer_imagenet_resnet_101_v3.py		infer_imagenet_resnet_101_v3.py
infer_imagenet_vgg_19.py		infer_imagenet_vgg_19.py
infer_little_image.py		infer_little_image.py
infer_little_image_15channels.py		infer_little_image_15channels.py
infer_little_image_5channels.py		infer_little_image_5channels.py
infer_little_image_fc_densenet.py		infer_little_image_fc_densenet.py
infer_little_image_fc_densenet_6channels.py		infer_little_image_fc_densenet_6channels.py
infer_little_image_fc_densenet_potsdam.py		infer_little_image_fc_densenet_potsdam.py
infer_little_image_potsdam.py		infer_little_image_potsdam.py
infer_numpy.py		infer_numpy.py
infer_numpy_potsdam.py		infer_numpy_potsdam.py
layers_fc_densenet.py		layers_fc_densenet.py
png2tif.py		png2tif.py
queue_pipeline.py		queue_pipeline.py
sample_image_15chan_v2.py		sample_image_15chan_v2.py
sampling_image.py		sampling_image.py
sampling_image_15channels.py		sampling_image_15channels.py
sampling_image_5channels.py		sampling_image_5channels.py
sampling_image_6channels.py		sampling_image_6channels.py
sampling_image_potsdam.py		sampling_image_potsdam.py

Repository files navigation

Student name: Nguyen Duc Minh
Class: CN-CNTT2 K59
Student ID: 20146488

Bachelor Final Project: Environment Setup and Tutorials

1. Environment Setup:
  - Python 3 Interpreter.
  - Python 3 libraries: OpenCV3, Tensorflow 1.4
  - Ubuntu 16.04.
2. Pretrained models: download them to dir named pretrained_models
  - VGG19: http://www.vlfeat.org/matconvnet/models/beta16/imagenet-vgg-verydeep-19.mat
  - ResNet101: http://www.vlfeat.org/matconvnet/models/imagenet-resnet-101-dag.mat
3. Tutorials:
  - Run demos:
    + $ python3 FCN-4s-ResNet101-6c-v3.3.py 0 (train FCN-4s-ResNet101-6c-v3.3 model)
    + $ python3 FCN-4s-ResNet101-5c-v3.3.py 0 (train FCN-4s-ResNet101-5c-v3.3 model)
    + $ python3 FCN-8s-ResNet101-5c-v3.2.py 0 (train FCN-8s-ResNet101-5c-v3.2 model)
    + $ python3 FCN-8s-ResNet101-5c-v3.1.py 0 (train FCN-8s-ResNet101-5c-v3.1 model)
    + $ python3 FCN-8s-ResNet101-5c-v3.0.py 0 (train FCN-8s-ResNet101-5c-v3.0 model)
  - Run with real datasets:
    + Request and download datasets from ISPRS 2D Labeling Contest Site.
    + Extract necessary data for each dataset (ISPRS_semantic_labeling_Vaihingen or
    ISPRS_semantic_labeling_Potsdam directory).
    + Vaihingen: $ python3 sampling_image_5channels.py
    + Potsdam: $ get_npy_Potsdam.py && sampling_image_potsdam_v4.py
4. Customisations:
  - Per patch accuracy evaluation: batch_eval_*.py, next_batch* methods in Batch_manager*

For detailed support and further collaboration: PLEASE contact me:
    - Email: gordonnguyen3796@gmail.com
    - Facebook: fb.me/minhnd3796
    - Tel: +84 167 633 1696