Embedded Files
DePondFi '23Ā
Detection of Pond Fish 2023 Challenge
Ā
The Problem Statement
The Problem Statement
Ā In this blue era, aquaculture plays a major role in our nationās economy. Fish disease is one of the critical issues that leads to huge productivity and economical loss in aquaculture farms ranging from small scale industries to large. Industry experts identify unhealthy fish through aberrant behaviour and it raises the labour load. Hence, aquaculture business needs a smart intelligence system to stop fish disease from spreading and nurturing fish wellness. Intelligent systems require automated and accurate detection of fishes in the underwater images/videos as the primary step for fish health monitoring. This challenge focuses on the first step (fish detection) for the development of automated pipelines in the process of fish disease detection and it must maintain a better trade-off between performance and time complexity. The detection of fish from underwater images is highly challenging owing to the complex and dynamic underwater environment with variable lighting conditions, water turbulence, and occlusion from other objects. Developing an effective fish detection algorithm can have a significant impact on the productivity of the aquaculture industry as it is the fundamental step in vision based fish health analysis.
Ā
Significance of the Challenge
Significance of the Challenge
āĀ In India, over 60% of the population consumes fish. The World Health Organization emphasises the importance of the health of fish because fish is consumed by the majority of the human population. Also, WHO recommends regular monitoring of fish health in the freshwater aqua farms to ensure food safety.
āĀ Early fish disease prediction through vision-based approach ensures the production of healthy and safe fish products. Consequently, it leads to sustainable aqua farms since it minimises usage of chemical-based antibiotics and labour load. Underwater image analysis is a non-invasive and effective method to monitor the fish heath in aquafarms. Fish detection in underwater images becomes the prime and challenging step in fish disease monitoring pipeline.
āĀ There are several competitions held around the world that focus on underwater image analysis: The Fishackathon, Aquatic Vision, ICPR, Ocean Discovery XPRIZE, Kaggle Fisheries Monitoring Competition. The most prominent events in the field of underwater image processing are the International Conference on UE, UASP, OMAE, OCEANS conference and so on.
Participation Rules
Participation Rules
āĀ Team size: Four members per team. A team lead should be there for further communication purposes.
āĀ The dataset will be made available only to the registered teams.
āĀ The challenge requires participants to develop a solution capable of detecting fish in real-time videos captured underwater with wild environment conditions in minimal time. The dataset will provide a detailed description and summary of each scenario, and annotations for the training set. Participants must submit the code for their proposed algorithm, written in Python and with appropriate comments included.
āĀ Teams are obliged to provide a summary of their approach and algorithms in a document/report as part of their submission. A demo script that allows the proposed solution to be run on a test video must also be included for further evaluation. Participants are also expected to share the inference time of their code, which is used as an evaluation metric, and provide information about the system specifications on which their code was implemented.
ā Ā Any malpractice will lead to disqualification of the entire team.
Evaluation Criteria
Evaluation Criteria
The algorithms proposed by the participants would be tested on various test cases of underwater image videos. mAP is the metric used for the evaluation process.
Mean Average Precision (mAP): This metric is commonly used in object detection tasks and measures the average precision across multiple thresholds for overlapping detections. It is calculated from precision-recall curve for different confidenceĀ thresholds, and then computing the area under the curve.Ā
This evaluation metric can be used to assess the performance of fish detection algorithm on DEPONDFIā23 Dataset captured under wild conditions. Also, compute the time taken to detect a fish in an image. Important note: The proposed algorithm that exhibits higher accuracy in cases when the fishes are extremely small, far off and with various distortions in the underwater videos will be given higher weightage.Ā
Dataset details
Dataset details
Upon the release date, registered participants will be provided with dataset through mail. The folder will contain:
Ā ā A collection of images of different types of Indian cultivated fish that have text annotations in YOLOV5 format.Ā
ā A document that contains a brief overview about different scenarios that are considered for image acquisition.Ā
ā The following table provides further details about the dataset.Ā
Submission Details
Submission Details
Each participating team must submit their solutions in the following format. Each team is required to email their solution in a zipped file to ncvpripg2023depondfi@gmail.com with the following specifications:
Ā Ā· File naming format: TeamName_DePondFi_NCVPRIPG2023
Ā Ā· Email Subject: DePondFi 2023 - [TEAM_NAME] Challenge Submission
The body of the mail shall include:
Team name
Team leader's name and email address
Rest of the team members
Executable/source code attached or download links
Ā Each zipped file must contain the following items:
1. Visual results for all the test frames with fish detected bounding boxes.
2.Ā Parameters/coordinates of bounding boxes as per YOLOV5 format (text file)
3. The executable/source code (python file) which should include trained models or necessary parameters so that we could run it and reproduce results. There should be a README or descriptions that explains how to execute theĀ executable/code
Prizes
Prizes
The top six teams will be invited for a presentation of their solution in a dedicated session at NCVPRIPG 2023. TA will be provided.
First Prize - Rs 5000
Second Prize - Rs 3000
Third Prize - Rs 2000
Our Sponsor
Our Sponsor
Schedule
Schedule
April 15, 2023
April 15, 2023
Registration opening and launch of challenge websiteĀ
Registration opening and launch of challenge websiteĀ
April 28, 2023
April 28, 2023
Release of training Data
Release of training Data
May 18, 2023
May 18, 2023
Availability of test Data
Availability of test Data
June 10, 2023
June 10, 2023
Submission ends
Submission ends
June 25, 2023
June 25, 2023
Announcement of Challenge Winners
Announcement of Challenge Winners
July 1, 2023
July 1, 2023
Registration Deadline for Challenge AttendeesĀ
Registration Deadline for Challenge AttendeesĀ
Organizers
Organizers
Dr Sasithradevi A
Dr Sasithradevi A
VIT Chennai
Dr R Suganya
Dr R Suganya
VIT Chennai
Dr P Prakash
Dr P Prakash
Anna Univ Chennai
Dr S Md Mansoor Roomi
Dr S Md Mansoor Roomi
TCE Madurai
Vijayalakshmi M
Vijayalakshmi M
VIT Chennai
Volunteers
Volunteers
Sabarinathan
Sabarinathan
Couger Inc., Japan
Persiya J
Persiya J
VIT Chennai
Kasthuri P
Kasthuri P
MIT Chennai
Brighty Ebenezer L
VIT Chennai
Brighty Ebenezer L
VIT Chennai
Akilesh S
Akilesh S
VIT Chennai
Tulasi Raman R
Tulasi Raman R
VIT Chennai
Page updated
Google Sites
Report abuse