Official Repository of NeurIPS2021 paper: PTR

Related tags

Deep LearningPTR
Overview

PTR: A Benchmark for Part-based Conceptual, Relational, and Physical Reasoning

Dataset Overview

Figure 1. Dataset Overview.

Introduction

A critical aspect of human visual perception is the ability to parse visual scenes into individual objects and further into object parts, forming part-whole hierarchies. Such composite structures could induce a rich set of semantic concepts and relations, thus playing an important role in the interpretation and organization of visual signals as well as for the generalization of visual perception and reasoning. However, existing visual reasoning benchmarks mostly focus on objects rather than parts. Visual reasoning based on the full part-whole hierarchy is much more challenging than object-centric reasoning due to finer-grained concepts, richer geometry relations, and more complex physics. Therefore, to better serve for part-based conceptual, relational and physical reasoning, we introduce a new large-scale diagnostic visual reasoning dataset named PTR. PTR contains around 70k RGBD synthetic images with ground truth object and part level annotations regarding semantic instance segmentation, color attributes, spatial and geometric relationships, and certain physical properties such as stability. These images are paired with 700k machine-generated questions covering various types of reasoning types, making them a good testbed for visual reasoning models. We examine several state-of-the-art visual reasoning models on this dataset and observe that they still make many surprising mistakes in situations where humans can easily infer the correct answer. We believe this dataset will open up new opportunities for part-based reasoning.

PTR is accepted by NeurIPS 2021.

Authors: Yining Hong, Li Yi, Joshua B Tenenbaum, Antonio Torralba and Chuang Gan from UCLA, MIT, IBM, Stanford and Tsinghua.

Arxiv Version: https://arxiv.org/abs/2112.05136

Project Page: http://ptr.csail.mit.edu/

Download

Data and evaluation server can be found here

TODOs

baseline models will be available soon!

About the Data

The data includes train/val/test images / questions / scene annotations / depths. Note that due to data cleaning process, the indices of the images are not necessarily consecutive.

The scene annotation is a json file that contains the following keys:

    cam_location        #location of the camera
    cam_rotation        #rotation of the camera
    directions          #Based on the camera, the vectors of the directions
    image_filename      #the filename of the image
    image_index         #the index of the image
    objects             #the objects in the scene, which contains a list of objects
        3d_coords       #the location of the object
        category        #the object category
        line_geo        #a dictionary containing (part, line unit normal vector) pairs. See the [unit normal vector](https://sites.math.washington.edu/~king/coursedir/m445w04/notes/vector/normals-plane.html) of a line. If the vector is not a unit vector, then the part cannot be considered a line.
        plane_geo       #a dictionary containing (part, plane unit normal vector) pairs. See the [unit normal vector](https://sites.math.washington.edu/~king/coursedir/m445w04/notes/vector/normals-plane.html) of a plane. If the vector is not a unit vector, then the part cannot be considered a line.
        obj_mask        #the mask of the object
        part_color      #a dictionary containing the colors of the parts
        part_count      #a dictionary containing the number of the parts
        part_mask       #a dictionary containing the masks of the parts
        partnet_id      #the id of the original partnet object in the PartNet dataset
        pixel_coords    #the pixel of the object
    relationships       #according to the directions, the spatial relationships of the objects
    projection_matrix   #the projection matrix of the camera to reconstruct 3D scene using depths
    physics(optional)   #if physics in the keys and the key is True, this is a physical scene.

The question file is a json file which contains a list of questions. Each question has the following keys:

    image_filename      #the image file that the question asks about
    image_index         #the image index that the question asks about
    program             #the original program used to generate the question
    program_nsclseq     #rearranged program as described in the paper
    question            #the question text
    answer              #the answer text
    type1               #the five questions types
    type2               #the 14 subtypes described in Table 2 in the paper

Data Generation Engine

The images and scene annotations can be generated via invoking data_generation/image_generation/render_images_partnet.py

blender --background --python render_images_partnet.py -- [args]

To generate physical scenes, invoke data_generation/image_generation/render_images_physics.py

blender --background --python render_images_physics.py -- [args]

For more instructions on image generation, please go to this directory and see the README file

To generate questions and answers based on the images, please go to this directory, and run

python generate_questions.py --input_scene_dir $INPUT_SCENE_DIR --output_dir $OUTPUT_QUESTION_DIR --output_questions_file $OUTPUT_FILE

The data generation engine is based partly on the CLEVR generation engine.

Errata

We have manually examined the images, annotations and questions twice. However, provided that there are annotation errors of the PartNet dataset we used, there could still be some errors in the scene annotations. If you find any errors that make the questions unanswerable, please contact [email protected].

Citations

@inproceedings{hong2021ptr,
author = {Hong, Yining and Yi, Li and Tenenbaum, Joshua B and Torralba, Antonio and Gan, Chuang},
title = {PTR: A Benchmark for Part-based Conceptual, Relational, and Physical Reasoning},
booktitle = {Advances In Neural Information Processing Systems},
year = {2021}
}
Owner
Yining Hong
https://evelinehong.github.io
Yining Hong
Code for paper " AdderNet: Do We Really Need Multiplications in Deep Learning?"

AdderNet: Do We Really Need Multiplications in Deep Learning? This code is a demo of CVPR 2020 paper AdderNet: Do We Really Need Multiplications in De

HUAWEI Noah's Ark Lab 915 Jan 01, 2023
This is the repo for the paper "Improving the Accuracy-Memory Trade-Off of Random Forests Via Leaf-Refinement".

Improving the Accuracy-Memory Trade-Off of Random Forests Via Leaf-Refinement This is the repository for the paper "Improving the Accuracy-Memory Trad

3 Dec 29, 2022
High level network definitions with pre-trained weights in TensorFlow

TensorNets High level network definitions with pre-trained weights in TensorFlow (tested with 2.1.0 = TF = 1.4.0). Guiding principles Applicability.

Taehoon Lee 1k Dec 13, 2022
Neural-PIL: Neural Pre-Integrated Lighting for Reflectance Decomposition - NeurIPS2021

Neural-PIL: Neural Pre-Integrated Lighting for Reflectance Decomposition Project Page | Video | Paper Implementation for Neural-PIL. A novel method wh

Computergraphics (University of Tübingen) 64 Dec 29, 2022
PyTorch version of the paper 'Enhanced Deep Residual Networks for Single Image Super-Resolution' (CVPRW 2017)

About PyTorch 1.2.0 Now the master branch supports PyTorch 1.2.0 by default. Due to the serious version problem (especially torch.utils.data.dataloade

Sanghyun Son 2.1k Jan 01, 2023
This is the source code for: Context-aware Entity Typing in Knowledge Graphs.

This is the source code for: Context-aware Entity Typing in Knowledge Graphs.

9 Sep 01, 2022
QAT(quantize aware training) for classification with MQBench

MQBench Quantization Aware Training with PyTorch I am using MQBench(Model Quantization Benchmark)(http://mqbench.tech/) to quantize the model for depl

Ling Zhang 29 Nov 18, 2022
WHENet: Real-time Fine-Grained Estimation for Wide Range Head Pose

WHENet: Real-time Fine-Grained Estimation for Wide Range Head Pose Yijun Zhou and James Gregson - BMVC2020 Abstract: We present an end-to-end head-pos

368 Dec 26, 2022
Unofficial Implementation of MLP-Mixer, Image Classification Model

MLP-Mixer Unoffical Implementation of MLP-Mixer, easy to use with terminal. Train and test easly. https://arxiv.org/abs/2105.01601 MLP-Mixer is an arc

Oğuzhan Ercan 6 Dec 05, 2022
Simultaneous NMT/MMT framework in PyTorch

This repository includes the codes, the experiment configurations and the scripts to prepare/download data for the Simultaneous Machine Translation wi

<a href=[email protected]"> 37 Sep 29, 2022
Doing the asl sign language classification on static images using graph neural networks.

SignLangGNN When GNNs 💜 MediaPipe. This is a starter project where I tried to implement some traditional image classification problem i.e. the ASL si

10 Nov 09, 2022
Implementation of MeMOT - Multi-Object Tracking with Memory - in Pytorch

MeMOT - Pytorch (wip) Implementation of MeMOT - Multi-Object Tracking with Memory - in Pytorch. This paper is just one in a line of work, but importan

Phil Wang 15 May 09, 2022
[CVPR 2021] Anycost GANs for Interactive Image Synthesis and Editing

Anycost GAN video | paper | website Anycost GANs for Interactive Image Synthesis and Editing Ji Lin, Richard Zhang, Frieder Ganz, Song Han, Jun-Yan Zh

MIT HAN Lab 726 Dec 28, 2022
A spherical CNN for weather forecasting

DeepSphere-Weather - Deep Learning on the sphere for weather/climate applications. The code in this repository provides a scalable and flexible framew

DeepSphere 47 Dec 25, 2022
Unified tracking framework with a single appearance model

Paper: Do different tracking tasks require different appearance model? [ArXiv] (comming soon) [Project Page] (comming soon) UniTrack is a simple and U

ZhongdaoWang 300 Dec 24, 2022
Python package to generate image embeddings with CLIP without PyTorch/TensorFlow

imgbeddings A Python package to generate embedding vectors from images, using OpenAI's robust CLIP model via Hugging Face transformers. These image em

Max Woolf 81 Jan 04, 2023
Official implementation of "Accelerating Reinforcement Learning with Learned Skill Priors", Pertsch et al., CoRL 2020

Accelerating Reinforcement Learning with Learned Skill Priors [Project Website] [Paper] Karl Pertsch1, Youngwoon Lee1, Joseph Lim1 1CLVR Lab, Universi

Cognitive Learning for Vision and Robotics (CLVR) lab @ USC 134 Dec 06, 2022
Gym-TORCS is the reinforcement learning (RL) environment in TORCS domain with OpenAI-gym-like interface.

Gym-TORCS Gym-TORCS is the reinforcement learning (RL) environment in TORCS domain with OpenAI-gym-like interface. TORCS is the open-rource realistic

naoto yoshida 400 Dec 27, 2022
PaSST: Efficient Training of Audio Transformers with Patchout

PaSST: Efficient Training of Audio Transformers with Patchout This is the implementation for Efficient Training of Audio Transformers with Patchout Pa

165 Dec 26, 2022
This is a TensorFlow implementation for C2-Rec

This is a TensorFlow implementation for C2-Rec We refer to the repo SASRec. Requirements requirement.txt Datasets This repo includes Amazon Beauty dat

7 Nov 14, 2022