Honours project, on creating a depth estimation map from two stereo images of featureless regions

Overview

image-processing

This module generates depth maps for shape-blocked-out images

Install

If working with anaconda, then from the root directory:

conda env create --file environment.yml
conda activate image-processing

Otherwise, if python 3 is installed, pip can be used to ensure the required packages are available. From the root directory, run

pip install -r requirements.txt

Files

The core functional files are collection.py, image.py, shape.py, edge.py, segment.py. They each contain a class of the same name. They logically follow this order and encapsulate each other, so collection creates three image objects for the left, center and right images. Each image object creates a number of shape objects. Shape objects create edge objects. Edge objects create segment objects. helper.py contains assisting functions used by these various classes.

This design aids in splitting up all the information and processes necessary to perform the desired function and logically groups it to ease comprehension. Each ought to be well-commented enough to generally understand what each part is doing.

The only one intended to be accessed to retrieve depth maps is collection.py as it orchestrates the entire process.

Usage

Both main.py and auto_gen.py are designed to access collection and to have it create depth maps. They require the initial images to be stored within a directory in assets/ , and each with three further subdirectories, cameraLeft/, cameraCenter/, and cameraRight/ . They save their results to saves/ with the generated images being stored in saves/generated/ . All .img files are object-files generated during this process to reduce the workload needed the next time the same process is executed.

Main

main.py is for individual depth map generation. There are four arguments able to be passed to specify details to the execution.

  1. The directory name desired from within assets/ .
  2. The numerical index (starting at 0) of the specific image desired within the innermost subdirectories
  3. The number representing which image should the depth map visual be based on (0 for left, 1 for center, 2 for right)
  4. Should the resulting depth image be saved
  5. Should the resulting depth image be displayed

While it can take up to these four arguments, no arguments is also possible. Then, the directory within assets/ is randomly selected, as is the index of the image set, and which image is used to generate the depth map visual. It will save and display the results. Partial arguments is also fine, so long as order is maintained.

Example: To display on the left image but not save occluded_road's first image set

python main.py occluded_road 0 0 False True

note: the last argument, True, is redundant in this case
 
Example: Any road_no_occlusion image set, any image used to create the depth map visual (automatically will save and display)

python main.py road_no_occlusion

 

Example: Anything (automatically will save and display)

python main.py

 

The value of having it execute a certain image when its depth image has already been generated is that it will quickly pull it up in the viewer and unlike the static image one can view the individual pixel values the mouse hovers over in the top-right corner.

 

Auto_gen

Alternatively, auto_gen.py is intended for the automated creation of all depth map images.

python auto_gen.py

By simply executing it, it will determine the depth map image all image sets and save them all. The terminal output is saved to a txt file stored in saves/logs. It does not display the results, as that would greatly heed the process of creating all of the results.

Alternatively, it can take two arguments.

  1. Specifies a directory within assets/ to use rather than executing for all of them, similar to the first argument for main.py
  2. Specifies the image to be used as the basis for the depth map visual, similar to the third argument for main.py (0-2 for left, center, and right)

Example: All depth images

python auto_gen.py

 
Example: All Shape_based_stereoPairs depth images using the right image

python auto_gen.py Shape_based_stereoPairs 2

 

For both, if an existing depth map exists, it will not be redone even if the image expected to be used is different. To do so, remove both the .jpg and .img and re-run.

How it works

 

Initialization

Upon creation of an instance of collection, it first intantiates the left image's Image intance. The shape colours are determined and then each shape is instantiated. The bounding box of the given shape is determined as well as its left and right edges, and their segments.

Collection uses the colours determined by the left Image to speed up the other two image's instantiations.

After everything has been created, the segments of each edge, of each shape, in each image must be assigned. First this process requires determining the displacement of edges, which is then used to determine which shape owns and doesn't own which segment.

Generally at this stage all but a few stragglers are assigned. The remaining are due to shapes having few edges, and the only one it could own is shared with the ground or sky shape, and thus difficult to tell which owns it. Using additional information about the shapes ownership is assigned. Finally, it checks to see if any shapes are the ground or sky, as their depths are not calculated.

At this stage, the image objects are saved.

Depth calculation

Then, using this information about the edges of a shape, its depth can be more accurately calculated. Only edges it owns are used to determine its depth. So if it only has its right side, only the right edge is used. Alternatively if both are owned, the midpoint is used.

However, if the shape is determined to have a varying depth, then its depth can alternatively be calculated using the change of slope between the images.

Finally, once all depth values are found, a modified version of the original image is created with its shape colours replaced with their determined depth values, the sky is replaced with pure black, and the ground with pure white. This image is then possibly saved and possibly displayed. Which image is used to re-colour for the depth map depends on either a given argument or random selection.

What can linearized neural networks actually say about generalization?

What can linearized neural networks actually say about generalization? This is the source code to reproduce the experiments of the NeurIPS 2021 paper

gortizji 11 Dec 09, 2022
2021 National Underwater Robotics Vision Optics

2021-National-Underwater-Robotics-Vision-Optics 2021年全国水下机器人算法大赛-光学赛道-B榜精度第18名 (Kilian_Di的团队:A榜[email pro

Di Chang 9 Nov 04, 2022
Semi-supervised Implicit Scene Completion from Sparse LiDAR

Semi-supervised Implicit Scene Completion from Sparse LiDAR Paper Created by Pengfei Li, Yongliang Shi, Tianyu Liu, Hao Zhao, Guyue Zhou and YA-QIN ZH

114 Nov 30, 2022
Benchmark for Answering Existential First Order Queries with Single Free Variable

EFO-1-QA Benchmark for First Order Query Estimation on Knowledge Graphs This repository contains an entire pipeline for the EFO-1-QA benchmark. EFO-1

HKUST-KnowComp 14 Oct 24, 2022
MEDS: Enhancing Memory Error Detection for Large-Scale Applications

MEDS: Enhancing Memory Error Detection for Large-Scale Applications Prerequisites cmake and clang Build MEDS supporting compiler $ make Build Using Do

Secomp Lab at Purdue University 34 Dec 14, 2022
Pytorch implementation of CVPR2020 paper “VectorNet: Encoding HD Maps and Agent Dynamics from Vectorized Representation”

VectorNet Re-implementation This is the unofficial pytorch implementation of CVPR2020 paper "VectorNet: Encoding HD Maps and Agent Dynamics from Vecto

120 Jan 06, 2023
Deep Learning and Reinforcement Learning Library for Scientists and Engineers 🔥

TensorLayer is a novel TensorFlow-based deep learning and reinforcement learning library designed for researchers and engineers. It provides an extens

TensorLayer Community 7.1k Dec 29, 2022
E-RAFT: Dense Optical Flow from Event Cameras

E-RAFT: Dense Optical Flow from Event Cameras This is the code for the paper E-RAFT: Dense Optical Flow from Event Cameras by Mathias Gehrig, Mario Mi

Robotics and Perception Group 71 Dec 12, 2022
Code for CoMatch: Semi-supervised Learning with Contrastive Graph Regularization

CoMatch: Semi-supervised Learning with Contrastive Graph Regularization (Salesforce Research) This is a PyTorch implementation of the CoMatch paper [B

Salesforce 107 Dec 14, 2022
Text Summarization - WCN — Weighted Contextual N-gram method for evaluation of Text Summarization

Text Summarization WCN — Weighted Contextual N-gram method for evaluation of Text Summarization In this project, I fine tune T5 model on Extreme Summa

Aditya Shah 1 Jan 03, 2022
🚀 An end-to-end ML applications using PyTorch, W&B, FastAPI, Docker, Streamlit and Heroku

🚀 An end-to-end ML applications using PyTorch, W&B, FastAPI, Docker, Streamlit and Heroku

Made With ML 82 Jun 26, 2022
Compartmental epidemic model to assess undocumented infections: applications to SARS-CoV-2 epidemics in Brazil - Datasets and Codes

Compartmental epidemic model to assess undocumented infections: applications to SARS-CoV-2 epidemics in Brazil - Datasets and Codes The codes for simu

1 Jan 12, 2022
This repo provides the base code for pytorch-lightning and weight and biases simultaneous integration.

Write your model faster with pytorch-lightning-wadb-code-backbone This repository provides the base code for pytorch-lightning and weight and biases s

9 Mar 29, 2022
Source code for CVPR 2020 paper "Learning to Forget for Meta-Learning"

L2F - Learning to Forget for Meta-Learning Sungyong Baik, Seokil Hong, Kyoung Mu Lee Source code for CVPR 2020 paper "Learning to Forget for Meta-Lear

Sungyong Baik 29 May 22, 2022
Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data

Real-ESRGAN Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Data Ported from https://github.com/xinntao/Real-ESRGAN Depend

Holy Wu 44 Dec 27, 2022
(Py)TOD: Tensor-based Outlier Detection, A General GPU-Accelerated Framework

(Py)TOD: Tensor-based Outlier Detection, A General GPU-Accelerated Framework Background: Outlier detection (OD) is a key data mining task for identify

Yue Zhao 127 Jan 05, 2023
Video Instance Segmentation using Inter-Frame Communication Transformers (NeurIPS 2021)

Video Instance Segmentation using Inter-Frame Communication Transformers (NeurIPS 2021) Paper Video Instance Segmentation using Inter-Frame Communicat

Sukjun Hwang 81 Dec 29, 2022
Neural network chess engine trained on Gary Kasparov's games.

Neural Chess It's not the best chess engine, but it is a chess engine. Proof of concept neural network chess engine (feed-forward multi-layer perceptr

3 Jun 22, 2022
Monocular Depth Estimation Using Laplacian Pyramid-Based Depth Residuals

LapDepth-release This repository is a Pytorch implementation of the paper "Monocular Depth Estimation Using Laplacian Pyramid-Based Depth Residuals" M

Minsoo Song 205 Dec 30, 2022
The Pytorch code of "Joint Distribution Matters: Deep Brownian Distance Covariance for Few-Shot Classification", CVPR 2022 (Oral).

DeepBDC for few-shot learning        Introduction In this repo, we provide the implementation of the following paper: "Joint Distribution Matters: Dee

FeiLong 116 Dec 19, 2022