ypwhs / Dogbreed_gluon
kaggle Dog Breed Identification
Stars: ✭ 116
Projects that are alternatives of or similar to Dogbreed gluon
Amazon Forest Computer Vision
Amazon Forest Computer Vision: Satellite Image tagging code using PyTorch / Keras with lots of PyTorch tricks
Stars: ✭ 346 (+198.28%)
Mutual labels: kaggle, jupyter-notebook, transfer-learning
Allstate capstone
Allstate Kaggle Competition ML Capstone Project
Stars: ✭ 72 (-37.93%)
Mutual labels: kaggle, jupyter-notebook
My Journey In The Data Science World
📢 Ready to learn or review your knowledge!
Stars: ✭ 1,175 (+912.93%)
Mutual labels: kaggle, jupyter-notebook
Weakly Supervised 3d Object Detection
Weakly Supervised 3D Object Detection from Point Clouds (VS3D), ACM MM 2020
Stars: ✭ 61 (-47.41%)
Mutual labels: jupyter-notebook, transfer-learning
Sru Deeplearning Workshop
دوره 12 ساعته یادگیری عمیق با چارچوب Keras
Stars: ✭ 66 (-43.1%)
Mutual labels: jupyter-notebook, transfer-learning
Imageclassification
Deep Learning: Image classification, feature visualization and transfer learning with Keras
Stars: ✭ 83 (-28.45%)
Mutual labels: jupyter-notebook, transfer-learning
Aws Machine Learning University Accelerated Cv
Machine Learning University: Accelerated Computer Vision Class
Stars: ✭ 1,068 (+820.69%)
Mutual labels: jupyter-notebook, gluon
Kaggle Competitions
There are plenty of courses and tutorials that can help you learn machine learning from scratch but here in GitHub, I want to solve some Kaggle competitions as a comprehensive workflow with python packages. After reading, you can use this workflow to solve other real problems and use it as a template.
Stars: ✭ 86 (-25.86%)
Mutual labels: kaggle, jupyter-notebook
Deep Learning Boot Camp
A community run, 5-day PyTorch Deep Learning Bootcamp
Stars: ✭ 1,270 (+994.83%)
Mutual labels: kaggle, jupyter-notebook
Tweet Stance Prediction
Applying NLP transfer learning techniques to predict Tweet stance
Stars: ✭ 91 (-21.55%)
Mutual labels: jupyter-notebook, transfer-learning
Neural Painters X
Neural Paiters
Stars: ✭ 61 (-47.41%)
Mutual labels: jupyter-notebook, transfer-learning
Fraud Detection
Credit Card Fraud Detection using ML: IEEE style paper + Jupyter Notebook
Stars: ✭ 58 (-50%)
Mutual labels: kaggle, jupyter-notebook
Tensorflow
This Repository contains all tensorflow tutorials.
Stars: ✭ 68 (-41.38%)
Mutual labels: jupyter-notebook, transfer-learning
Ko en neural machine translation
Korean English NMT(Neural Machine Translation) with Gluon
Stars: ✭ 55 (-52.59%)
Mutual labels: jupyter-notebook, gluon
Kaggle Notebooks
Sample notebooks for Kaggle competitions
Stars: ✭ 77 (-33.62%)
Mutual labels: kaggle, jupyter-notebook
Dog Breeds Classification
Set of scripts and data for reproducing dog breed classification model training, analysis, and inference.
Stars: ✭ 105 (-9.48%)
Mutual labels: kaggle, jupyter-notebook
Lung Diseases Classifier
Diseases Detection from NIH Chest X-ray data
Stars: ✭ 52 (-55.17%)
Mutual labels: kaggle, jupyter-notebook
Average Word2vec
🔤 Calculate average word embeddings (word2vec) from documents for transfer learning
Stars: ✭ 52 (-55.17%)
Mutual labels: jupyter-notebook, transfer-learning
Intrusion Detection System Using Deep Learning
VGG-19 deep learning model trained using ISCX 2012 IDS Dataset
Stars: ✭ 85 (-26.72%)
Mutual labels: jupyter-notebook, transfer-learning
使用 gluon 实现狗品种分类
本项目是为了熟悉 gluon 操作的练手作,可以在测试集上获得0.28228的分数。
环境参考:
- mxnet 0.12.1
- numpy 1.13.3
- tqdm 4.11.2
- pandas 0.20.3
- sklearn 0.19.0
思路
复现步骤分为以下几步:
- 下载并解压数据集:https://www.kaggle.com/c/dog-breed-identification/data
- 运行 preprocessing.ipynb,构建 for_train 和 for_test 的文件夹结构,为 ImageFolderDataset 做准备
- 运行 get_features.ipynb,导出 resnet152_v1, densenet161 和 inception_v3 等模型输出的特征,参考model_zoo
- 运行 model.ipynb,拼接特征,训练一个2层的神经网络,然后在测试集上进行预测
Update
为了确定使用的预训练模型是否是最好的,我对所有预训练模型进行了迁移学习:
- get_features_v3.ipynb 导出所有预训练模型输出的特征
- transfer_learning_all_pretrained_models.ipynb 对所有预训练模型分别进行迁移学习,按 val_loss 进行排序
model | val_loss |
---|---|
inceptionv3 | 0.296050225385 |
resnet152_v1 | 0.399359531701 |
resnet101_v1 | 0.410383010283 |
densenet161 | 0.418100789189 |
densenet201 | 0.453403010964 |
resnet50_v2 | 0.484435886145 |
resnet50_v1 | 0.496179759502 |
densenet169 | 0.512498702854 |
resnet34_v2 | 0.536734519526 |
vgg19_bn | 0.557294445112 |
vgg16_bn | 0.586511127651 |
resnet34_v1 | 0.591432901099 |
densenet121 | 0.591716498137 |
vgg19 | 0.619780953974 |
vgg16 | 0.669267293066 |
vgg13_bn | 0.702507363632 |
vgg11_bn | 0.708396691829 |
vgg13 | 0.756541173905 |
resnet18_v2 | 0.761708110571 |
vgg11 | 0.789955694228 |
resnet18_v1 | 0.832537706941 |
squeezenet1.1 | 1.6066500321 |
squeezenet1.0 | 1.62178872526 |
alexnet | 1.77026221156 |
可以看到 densenet 并没有那么好,于是我只使用 inceptionv3 和 resnet152_v1 的特征,进行了融合迁移学习,获得了 0.27143 的分数。
Update
使用 Stanford Dogs Dataset 进行训练,使用了 inceptionv3 和 resnet152_v1,最终获得了 0.00398 的分数。
- get_features_v3.ipynb 导出所有预训练模型输出的特征
- stanford.ipynb 导出 Stanford Dogs Dataset 对应的特征,然后进行迁移学习,最后在测试集上进行预测
Note that the project description data, including the texts, logos, images, and/or trademarks,
for each open source project belongs to its rightful owner.
If you wish to add or remove any projects, please contact us at [email protected].