Deep Learning::Coisas Prontas & Datasets

Redes Prontas e Treinadas

Compilações

1. Discussão interessante: Top 10 Pretrained Models to get you Started with Deep Learning (Part 1 – Computer Vision) – Indica vários modelos e paltaformas, sem focar em nenhum.

Keras/Tensor Flow

1. Keras implementation of RetinaNet object detection – https://github.com/fizyr/keras-retinanet

PyTorch

1. Awesome-Pytorch-list (Redes, Tutoriais, Extensões e Bibliotecas) – https://github.com/bharathgs/Awesome-pytorch-list
2. Pretrained ConvNets for pytorch: NASNet, ResNeXt, ResNet, InceptionV4, InceptionResnetV2, Xception, DPN, etc. – https://github.com/Cadene/pretrained-models.pytorch
3. PyTorch – pretrained torchvision and Emotion Recognition examples – http://www.robots.ox.ac.uk/~albanie/pytorch-models.html
4. https://github.com/pytorch/vision/tree/master/torchvision/models

DNN (OpenCV)

1.  nada ainda…

Outros

1. Caffe Model Zoo – http://caffe.berkeleyvision.org/model_zoo.html

Imagens

1. ImageNet Tree View
2. COCO – Common Objects in Context
3. The PASCAL Visual Object Classes
4. Kitti Road dataset com download daqui
5. Google Open Images Dataset V4
6. Photo Aesthetics Ranking Network with Attributes and Content Adaptation
7. AVA: A Large-Scale Database for Aesthetic Visual Analysis
8. Stanford Background Dataset
9. Sift Flow Dataset
10. Barcelona Dataset
11. Microsoft COCO dataset
12. MSRC Dataset
13. LITS Liver Tumor Segmentation Dataset
14. Data from Games dataset
15. Human parsing dataset
16. Mapillary Vistas Dataset
17. Microsoft AirSim
18. MIT Scene Parsing Benchmark
19. ADE20K Dataset
20. INRIA Annotations for Graz-02
21. Daimler dataset
22. ISBI Challenge: Segmentation of neuronal structures in EM stacks
23. INRIA Annotations for Graz-02 (IG02)
24. Pratheepan Dataset
25. Clothing Co-Parsing (CCP) Dataset
26. Inria Aerial Image
27. ApolloScape
28. UrbanMapper3D
29. RoadDetector

Fontes de Video

1. Passagem de pessoas rua  – http://www.cxtv.com.br/camera-ao-vivo/temple-bar
2. Distante  – http://www.cxtv.com.br/camera-ao-vivo/mashhad
3. Mercado publico de Florianópolis, lateral e mesas  – http://www.pmf.sc.gov.br/temporeal/
4. Mirante da Lagoa da Conceição – http://maaxcam.com.br/
5. Loja Japonesa – https://www.insecam.org/en/view/690554/   e  http://www.svcl.ucsd.edu/projects/peoplecnt/

Datasets Gerais

1. Medium::The 50 Best Public Datasets for Machine Learning

Bases para Benchmarks e Competições

ImageNet

The ImageNet project is a large visual database designed for use in visual object recognition software research. The ImageNet project runs an annual software contest, the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), where software programs compete to correctly classify and detect objects and scenes.

 

PASCAL VOC – Visual Object Classes

The PASCAL VOC project: http://host.robots.ox.ac.uk/pascal/VOC/

• Provides standardised image data sets for object class recognition
• Provides a common set of tools for accessing the data sets and annotations
• Enables evaluation and comparison of different methods 
• Ran challenges evaluating performance on object class recognition (from 2005-2012, now finished)
• Leaderboards for the Evaluations on PASCAL VOC Data

Pascal VOC data sets: Data sets from the VOC challenges are available through the challenge links below, and evalution of new methods on these data sets can be achieved through the  PASCAL VOC Evaluation Server.  The evaluation server will remain active even though the challenges have now finished.

Microsoft COCO – Common Objects in COntext

COCO is a large-scale object detection, segmentation, and captioning dataset. COCO has several features: Object segmentation, Recognition in context, Superpixel stuff segmentation, 330K images (>200K labeled), 1.5 million object instances, 80 object categories, 91 stuff categories, 5 captions per image and 250,000 people with keypoints.

COCO:

• Site: http://cocodataset.org
• Artigo: Microsoft COCO: Common Objects in Context – Tsung-Yi Lin, Michael Maire, Serge Belongie, Lubomir Bourdev, Ross Girshick, James Hays, Pietro Perona, Deva Ramanan, C. Lawrence Zitnick, Piotr Dollár
• GitHub: https://github.com/nightrome/cocostuff

CIFAR-10 e CIFAR-100

CIFAR-10 e CIFAR-100 são subconjuntos anotados do 80 million tiny images dataset. Foram coletados por Alex Krizhevsky, Vinod Nair e Geoffrey Hinton.É usado como referência para treinamento de muitos modelos de CNNs.

• Site: https://www.cs.toronto.edu/~kriz/cifar.html
• Competição no Kaggle: CIFAR-10 – Object Recognition in Images
• Tutorial: Towards Data Science – CIFAR-10 Image Classification in TensorFlow
• Github: https://github.com/EN10/CIFAR

Google Open Images Dataset V4

• https://storage.googleapis.com/openimages/web/index.html
• Open Images Challenge 2018 – (dataset)

Kaggle

Kaggle é uma comunidade internacional de cientistas de dados e pesquisadores em Aprendizado de Máquina, famosa por suas competições de aprendizado de máquina, onde problemas específicos com datasets associados são postados e times de diferentes instituições e empresas competem para tentar resolver. Passou também a oferecer uma plataforma pública para armazenamento de datasets e um ambiente de oficina de aprendizado de máquina baseado em nuvem. Me Março de 2017 Google anunciou a aquisição de Kaggle, que continua a operar nos antigos moldes.

• Site: https://www.kaggle.com/
• Competições: this was Kaggle’s first product and still what the site is most famous for. Companies post problems and machine learners compete to build the best algorithm. In addition, Kaggle also has:
• Kaggle Kernels: a cloud-based workbench for data science and machine learning. Allows data scientists to shore code and analysis in Python and R. Over 150K “kernels” (code snippets) have been shared on Kaggle covering everything from sentiment analysis to object detection.
• Datasets: community members share datasets with each other. Has datasets on everything from bone x-rays to results from boxing bouts.
• Kaggle Learn: cursos de IA de curta duração.

Alibabab Tianchi

Plataforma de Ciência de Dados e Aprendizado de Máquina do provedor de e-commerce Alibaba. Similar ao Kaggle, oferece competições, plataforma de deep learning, datasets, cursos e oportunidades de emprego.

• Site: https://tianchi.aliyun.com/
• Competições: https://tianchi.aliyun.com/competition/gameList.htm
• Plataforma de aprendizado de máquina: https://tianchi.aliyun.com/notebook/index.htm

InnoCentive

Empresa de inovação de crowdsourcing que também oferece competições e desafios similares ao Kaggle.

• Site: https://www.innocentive.com/
• Desafios/competições: https://www.innocentive.com/ar/challenge/browse

Copyright © 2018 Aldo von Wangenheim/INCoD/Universidade Federal de Santa Catarina