Mini Places Challenge Adrià Recasens, Nov 21.

Slides:

Advertisements

Similar presentations

Rich feature Hierarchies for Accurate object detection and semantic segmentation Ross Girshick, Jeff Donahue, Trevor Darrell, Jitandra Malik (UC Berkeley)

Advertisements

ImageNet Classification with Deep Convolutional Neural Networks

Neural Networks Chapter Feed-Forward Neural Networks.

Spatial Pyramid Pooling in Deep Convolutional

What is the Best Multi-Stage Architecture for Object Recognition Kevin Jarrett, Koray Kavukcuoglu, Marc’ Aurelio Ranzato and Yann LeCun Presented by Lingbo.

Convolutional Neural Networks for Image Processing with Applications in Mobile Robotics By, Sruthi Moola.

Dr. Z. R. Ghassabi Spring 2015 Deep learning for Human action Recognition 1.

Object detection, deep learning, and R-CNNs

Feedforward semantic segmentation with zoom-out features

Notes on HW 1 grading I gave full credit as long as you gave a description, confusion matrix, and working code Many people’s descriptions were quite short.

ImageNet Classification with Deep Convolutional Neural Networks Presenter: Weicong Chen.

Object Recognizing. Deep Learning Success in 2012 DeepNet and speech processing.

Deep Residual Learning for Image Recognition

Deep Learning Overview Sources: workshop-tutorial-final.pdf

Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition arXiv: v4 [cs.CV(CVPR)] 23 Apr 2015 Kaiming He, Xiangyu Zhang, Shaoqing.

Assignment 4: Deep Convolutional Neural Networks

Facial Detection via Convolutional Neural Network Nathan Schneider.

Combining Models Foundations of Algorithms and Machine Learning (CS60020), IIT KGP, 2017: Indrajit Bhattacharya.

Automatic Grading of Diabetic Retinopathy through Deep Learning

CS 4501: Introduction to Computer Vision Object Localization, Detection, Semantic Segmentation Connelly Barnes Some slides from Fei-Fei Li / Andrej Karpathy.

CNN-RNN: A Uniﬁed Framework for Multi-label Image Classiﬁcation

The Relationship between Deep Learning and Brain Function

CS 6501: 3D Reconstruction and Understanding Convolutional Neural Networks Connelly Barnes.

Compact Bilinear Pooling

Computer Science and Engineering, Seoul National University

Convolutional Neural Fabrics by Shreyas Saxena, Jakob Verbeek

The Problem: Classification

Understanding and Predicting Image Memorability at a Large Scale

Applications of Deep Learning and how to get started with implementation of deep learning Presentation By : Manaswi Advisor : Dr.Chinmay.

Presenter: Chu-Song Chen

Inception and Residual Architecture in Deep Convolutional Networks

ECE 6504 Deep Learning for Perception

Predict House Sales Price

CAMELYON16 Challenge Matt Berseth, NLP Logix Jacksonville FL

Training Techniques for Deep Neural Networks

Efficient Deep Model for Monocular Road Segmentation

Deep Belief Networks Psychology 209 February 22, 2013.

A brief introduction to neural network

ECE 599/692 – Deep Learning Lecture 6 – CNN: The Variants

State-of-the-art face recognition systems

Bird-species Recognition Using Convolutional Neural Network

Introduction to Neural Networks

Image Classification.

CS 4501: Introduction to Computer Vision Training Neural Networks II

Logistic Regression & Parallel SGD

Construct a Convolutional Neural Network with Python

cs540 - Fall 2016 (Shavlik©), Lecture 20, Week 11

Road Traffic Sign Recognition

Very Deep Convolutional Networks for Large-Scale Image Recognition

Chap. 7 Regularization for Deep Learning (7.8~7.12 )

network of simple neuron-like computing elements

CSC 578 Neural Networks and Deep Learning

Neural Networks Geoff Hulten.

On Convolutional Neural Network

Visualizing and Understanding Convolutional Networks

Overview of Neural Network Architecture Assignment Code

Dr. Borji Aisha Urooj Cecilia La Place

Forward and Backward Max Pooling

Object Tracking: Comparison of

Neural Networks II Chen Gao Virginia Tech ECE-5424G / CS-5824

ImageNet Classification with Deep Convolutional Neural Networks

边缘检测年度进展概述 Ming-Ming Cheng Media Computing Lab, Nankai University

Heterogeneous convolutional neural networks for visual recognition

Neural Networks II Chen Gao Virginia Tech ECE-5424G / CS-5824

Department of Computer Science Ben-Gurion University of the Negev

Natalie Lang Tomer Malach

VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITION

Example of training and deployment of deep convolutional neural networks. Example of training and deployment of deep convolutional neural networks. During.

CSC 578 Neural Networks and Deep Learning

Principles of Back-Propagation

Presentation transcript:

Mini Places Challenge Adrià Recasens, Nov 21

Scene classification River

Scene classification Office

Project: Mini Places Challenge Goal: Build the best classiﬁer you can for scenes

The evolution of vision databases COILV20 2 year old kid Caltech 101 ObjectDcentered datasets MNIST (1998) (2009) 103 CaltechV4 (2003) 104 PASCAL (2005) 105 106 107 108 109 # images SceneDcentered datasets 15 scenes database (2006) SUN database (2010) 8 scenes database (2001) 103 104 105 106 107 108 109 # images

Start with 60 million images Computation with millions of pictures of places, scenes, environments – Where you and objects are Start with 60 million images

Amazon Mechanical Turk: Single Image Classiﬁcation

Amazon Mechanical Turk: Single Image Classiﬁcation

Nov 2015: Top 5 error: 25%

Convolution Max pooling Normalization Fully connected Convolutional Neural Network

More semantically meaningful

Challenge data Training data: 100K, 1000 images per category. Validation data: 10K, 100 images per category. Test data: 10K, 100 images per category.

Provided code data: categories, train/val, object categories. evaluation: evaluation code. util: formatting functions.

Challenge details Size: 128x128. Jpeg format. Object annotations for a subset of the images. 3503 train images and 371 validation images Evaluation server opens Monday Nov 21. Open until Dec 7. Errors with 5 guesses will be used to rank the results and determine the winner

Possible improvements Train different types of classifiers: objects and scenes to improve scene classification Ensemble methods: combine outputs from various networks trained with different random seeds and/or also networks with different architectures) Try averaging the predictions from multiple crops

Possible improvements Different architectures: numbers of units per layer, types of activation functions, supervision to intermediate layers. Different learning parameters: learning rate, weight decay, momentum and so on.

Infrastructure: Amazon EC2 AWS Education project: approx. 6 days of free GPU use per member of the group (mit account). Close the instance when you are done (time costs money!) Instance type: g2.2xlarge GitHub Education Pack: extra $50

Data loading Torch Data Loading packages Torch-starter: https://github.com/cvondrick/torch-starter Torch-proj: https://github.com/nhynes/torch-proj