Skip to content
Snippets Groups Projects
Commit e0c65faf authored by eric_a_scuccimarra's avatar eric_a_scuccimarra
Browse files

Update README.md

parent 60c88ab5
Branches main
No related tags found
No related merge requests found
......@@ -13,264 +13,9 @@ This is an evolving dataset, where we will release more data as the dataset grow
# Baseline
MMdetection is an open source object detection toolbox based on PyTorch, with a large Model Zoo with many customised models that can be plugged and tested in with just a single config file modification. PYou can read more about it at: [mmdetection github](https://github.com/open-mmlab/mmdetection/)
# Code
# Installation
[MMDetection Installation instructions](https://github.com/open-mmlab/mmdetection/blob/master/docs/INSTALL.md)
Follow the installation instructions as given in the above link.
Ensure you have `docker` and `nvidia-docker` installed by following the instructions here :
* [Docker](https://docs.docker.com/install/)
* [nvidia-docker](https://github.com/NVIDIA/nvidia-docker)
**NOTE** : You do not need nvidia-docker if you do not want to use GPU when testing your submission locally
```
git clone http://gitlab.aicrowd.com/nikhil_rayaprolu/food-pytorch-baseline.git
cd food-pytorch-baseline
pip3 install -r requirements.txt
pip3 install git+https://github.com/AIcrowd/coco.git
```
# Dataset
The dataset for the [AIcrowd Food Recognition Challenge](https://www.aicrowd.com/challenges/food-recognition-challenge) is available at [https://www.aicrowd.com/challenges/food-recognition-challenge/dataset_files](https://www.aicrowd.com/challenges/food-recognition-challenge/dataset_files)
This dataset contains :
* `train-v0.2.tar.gz` : This is the Training Set of **7949** (as RGB images) food images, along with their corresponding annotations in [MS-COCO format](http://cocodataset.org/#home)
* `val-v0.2.tar.gz`: This is the suggested Validation Set of **418** (as RGB images) food images, along with their corresponding annotations in [MS-COCO format](http://cocodataset.org/#home)
* `test_images-v0.2.tar.gz` : This is the debug Test Set for Round-1, where you are provided the same images as the validation set.
To get started, we would advise you to download all the files, and untar them inside the `data/` folder of this repository, so that you have a directory structure like this :
```bash
|-- data/
| |-- test_images/ (has all images for prediction)(**NOTE** : They are the same as the validation set images)
| |-- train/
| | |-- images (has all the images for training)
| | |__ annotation.json : Annotation of the data in MS COCO format
| | |__ annotation-small.json : Smaller version of the previous dataset
| |-- val/
| | |-- images (has all the images for training)
| | |__ annotation.json : Annotation of the data in MS COCO format
| | |__ annotation-small.json : Smaller version of the previous dataset
```
We are also assuming that you have already installed all the requirements for this notebook, or you can still install them by :
# Usage
# Training with MMDetection:
Let us look at training MMDetection using Hybrid Task Cascade [HTC research paper](https://arxiv.org/abs/1901.07518).
A score of AP_50 of 0.526 and AR_50 of 0.729, can be achieved with Hybrid Task Cascade of Resnet50 Backbone.
MMDetection provides us with a config file especially for HTC, available at [HTC config](https://github.com/open-mmlab/mmdetection/tree/master/configs/htc)
Also make sure you have downloaded the training data to a subfolder of your project.
Modify your config file and point your dataset variables to your data folder.
As given in [MMDetection Getting Started](https://github.com/open-mmlab/mmdetection/blob/master/docs/GETTING_STARTED.md),
You can use:
python tools/train.py ${CONFIG_FILE}
to train the model on a single GPU or
./tools/dist_train.sh ${CONFIG_FILE} ${GPU_NUM} [optional arguments]
to train the model on multiple GPUs.
Make sure you have edited the config file to point to the dataset and also have made changes to the number of classes.
If you are going to use the dataloader from the mmdetection.
## Testing with MMDetection:
To test your results with MMDetection,
you can use the commands:
```
*single-gpu testing*
python tools/test.py ${CONFIG_FILE} ${CHECKPOINT_FILE} [--out ${RESULT_FILE}] [--eval ${EVAL_METRICS}] [--show]
*multi-gpu testing*
./tools/dist_test.sh ${CONFIG_FILE} ${CHECKPOINT_FILE} ${GPU_NUM} [--out ${RESULT_FILE}] [--eval ${EVAL_METRICS}]
```
**Log Analysis**
The training logs can be analyzed using the plot_curve provided with the mmdetection:
```
import os
import matplotlib
%matplotlib inline
from tools.analyze_logs import plot_curve
matplotlib.rcParams['figure.figsize'] = [20, 10]
args = {
'keys':['segm_mAP_50'],
'legend':'segm_mAP_50',
'backend': None,
'json_logs': [os.getcwd()+'/work_dirs/htc_r50_fpn/20191206_105437.log.json'],
'title': 'loss'
}
print(os.getcwd()+'/work_dirs/htc_r50_fpn/20191206_105437.log.json')
plot_curve([os.getcwd()+'/work_dirs/htc_r50_fpn/20191206_105437.log.json'], args)
```
## Other Associated Notebooks
* [Dataset Utils](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/Dataset%20Utils.ipynb)
* [Import Dependencies](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/Dataset%20Utils.ipynb#Import-dependencies)
* [Configuration Variables](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/Dataset%20Utils.ipynb#Configuration-Variables)
* [Parsing Annotations](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/Dataset%20Utils.ipynb#Parsing-the-annotations)
* [Collecting and Visualizing Images](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/Dataset%20Utils.ipynb#Collecting-and-Visualizing-Images)
* [Understanding Annotations](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/Dataset%20Utils.ipynb#Understanding-Annotations)
* [Visualizing Annotations](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/Dataset%20Utils.ipynb#Visualizing-Annotations)
* [Advanced](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/Dataset%20Utils.ipynb#Advanced)
* [Convert poly segmentation to rle](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/Dataset%20Utils.ipynb#1.-Convert-poly-segmentation-to-rle)
* [Convert segmentation to pixel level masks](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/Dataset%20Utils.ipynb#2.-Convert-segmentation-to-pixel-level-masks)
* [Random Submission](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/run.py)
* [Locally test the evaluation function](https://github.com/AIcrowd/food-recognition-challenge-starter-kit/blob/master/Local%20Evaluation.ipynb)
## Other Baselines
# Round 1
* [Colab Notebook for Data Analysis and Tutorial](https://colab.research.google.com/drive/1A5p9GX5X3n6OMtLjfhnH6Oeq13tWNtFO#scrollTo=ok54AWT_VoWV)
A notebook with data analysis on the Food Recognition Dataset and then a short tutorial on training with keras and pytorch. This lets you immediately jump onto the challenge and solve the challenge
### Pretrained Baselines
* [mmdetection (pytorch)](https://gitlab.aicrowd.com/nikhil_rayaprolu/food-pytorch-baseline)
* [matterport-maskrcnn (keras - tensorflow)](https://gitlab.aicrowd.com/nikhil_rayaprolu/food-recognition)
# Round 2
* [Colab Notebook for Data Analysis and Tutorial](https://colab.research.google.com/drive/1vXdv9quZ7CXO5lLCjhyz3jtejRzDq221)
A notebook with data analysis on the Food Recognition Dataset and then a short tutorial on training with keras and pytorch. This lets you immediately jump onto the challenge and solve the challenge
### Pretrained Baselines
* [mmdetection (pytorch)](https://gitlab.aicrowd.com/nikhil_rayaprolu/food-round2)
# Submission Instructions
To submit to the challenge you'll need to ensure you've set up an appropriate repository structure, create a private git repository at https://gitlab.aicrowd.com with the contents of your submission, and push a git tag corresponding to the version of your repository you'd like to submit.
## Repository Structure
We have created this sample submission repository which you can use as reference.
#### aicrowd.json
Each repository should have a aicrowd.json file with the following fields:
```
{
"challenge_id" : "aicrowd-food-recognition-challenge",
"grader_id": "aicrowd-food-recognition-challenge",
"authors" : ["aicrowd-user"],
"description" : "Food Recognition Challenge Submission",
"license" : "MIT",
"gpu": true
}
```
This file is used to identify your submission as a part of the Food Recognition Challenge. You must use the `challenge_id` and `grader_id` specified above in the submission. The `gpu` key in the `aicrowd.json` lets your specify if your submission requires a GPU or not. In which case, a NVIDIA-K80 will be made available to your submission when evaluation the submission.
#### Submission environment configuration
You can specify the software runtime of your code by modifying the included [Dockerfile](Dockerfile).
#### Code Entrypoint
The evaluator will use `/home/aicrowd/run.sh` as the entrypoint. Please remember to have a `run.sh` at the root which can instantiate any necessary environment variables and execute your code. This repository includes a sample `run.sh` file.
### Local Debug
```
export TEST_IMAGES_PATH="../data/test_images" # or path to your local folder containing images
export IMAGE_NAME="aicrowd-food-recognition-challenge-submission"
./build.sh
./debug.sh
######################################
## NOTE :
##
## * If you do not wish to your a GPU when testing locally, please feel free to replace nvidia-docker with docker
##
## * If you want to test on images located at an alternate location, set the `TEST_IMAGES_PATH` environment variable accordingly before running `build.sh` and `debug.sh`.
```
### Submitting
To make a submission, you will have to create a private repository on [https://gitlab.aicrowd.com](https://gitlab.aicrowd.com).
You will have to add your SSH Keys to your GitLab account by following the instructions [here](https://docs.gitlab.com/ee/gitlab-basics/create-your-ssh-keys.html).
If you do not have SSH Keys, you will first need to [generate one](https://docs.gitlab.com/ee/ssh/README.html#generating-a-new-ssh-key-pair).
Then you can create a submission by making a *tag push* to your repository, adding the correct git remote and pushing to the remote:
```
git clone https://gitlab.aicrowd.com/nikhil_rayaprolu/food-pytorch-baseline
cd food-pytorch-baseline
# Add AICrowd git remote endpoint
git remote add aicrowd git@gitlab.aicrowd.com:<YOUR_AICROWD_USER_NAME>/food-challenge-pytorch-baseline.git
git push aicrowd master
# Create a tag for your submission and push
git tag -am "submission-v0.1" submission-v0.1
git push aicrowd master
git push aicrowd submission-v0.1
# Note : If the contents of your repository (latest commit hash) does not change,
# then pushing a new tag will not trigger a new evaluation.
```
You now should be able to see the details of your submission at :
[gitlab.aicrowd.com/<YOUR_AICROWD_USER_NAME>/food-challenge-pytorch-baseline/issues](gitlab.aicrowd.com/<YOUR_AICROWD_USER_NAME>/food-challenge-pytorch-baseline/issues)
## Using http instead of ssh (Personal Access Token):
In order to use http to clone repositories and submit on gitlab:
a) Create a personal access token
1. Log in to GitLab.
2. In the upper-right corner, click your avatar and select Settings.
3. On the User Settings menu, select Access Tokens.
4. Choose a name and optional expiry date for the token.
5. Choose the desired scopes.
6. Click the Create personal access token button.
7. Save the personal access token somewhere safe, lets call it XXX for now.
Once you leave or refresh the page, you won’t be able to access it again.
b) to clone a repo use the following command:
git clone [https://oauth2:XXX@gitlab.aicrowd.com/(username)/(repo_name).git](https://oauth2:XXX@gitlab.aicrowd.com/(username)/(repo_name).git)
c)submit a solution:
```
cd into your submission repo on gitlab
cd (repo_name)
#Add AICrowd git remote endpoint
git remote add aicrowd https://oauth2:XXX@gitlab.aicrowd.com/(username)/(repo_name).git
git push aicrowd master
# Create a tag for your submission and push
git tag -am "submission-v0.1" submission-v0.1
git push aicrowd master
git push aicrowd submission-v0.1
# Note : If the contents of your repository (latest commit hash) does not change,
# then pushing a new tag will not trigger a new evaluation.
```
This repo contains the source code used to train my best submission for Round 4. The submission was a two-model ensemble of DetectoRS and HTC x101.
**Best of Luck**
......@@ -283,5 +28,5 @@ git push aicrowd submission-v0.1
* and the baseline is built using MMDetection: https://github.com/open-mmlab/mmdetection/
# Author
**[Nikhil Rayaprolu](nikhil@aicrowd.com)**
**[Eric Antoine Scuccimarra](skooch@gmail.com)**
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment