The aim of the Machine Vision Innovation (MVI) Tournament is to foster recent advances in image processing, computer vision, and machine learning to solve different open-ended real-world challenges. The participants will be solving one of the four problems described below. The best three teams will be awarded a cash prize, whereas all participating teams will be given certificates.

1. Problems:

There are four different problems within the tournament. The participants can choose one of them to solve within the prescribed time frame. The summary of each problem is presented below, whereas their detailed descriptions are uploaded on the tournament website:

Problem-1: Detecting and Predicting UAV Sensory Failures through Advanced Image Processing and Machine Learning

Unmanned Aerial Vehicles (UAVs) play a crucial role in various applications, including surveillance, agriculture, and disaster response. The reliability and performance of UAVs heavily depend on the accurate functioning of their sensors. Detecting and predicting sensory failures in real-time is paramount for ensuring the safe and efficient operation of these UAVs. This problem is aimed to address the issue of UAV sensory failures through the utilization of state-of-the-art image processing, and machine learning techniques.

Problem-2: Detecting Structural Cracks in Buildings and Construction Sites

Ensuring the structural integrity and safety of buildings and construction sites is paramount in urban development. This problem is aimed to address this critical aspect by detecting structural cracks. Participants will be provided with a unique dataset that integrates visual data collected from various construction scenarios. The challenge will require participants to design innovative solutions using state-of-the-art image processing, computer vision, and machine learning schemes.

Problem-3: Crops Recognition Using Visual Tactile Information

Precision agriculture is at the forefront of modern farming practices, and the accurate assessment of crops condition is crucial for optimizing yields. This problem is aimed to address this agricultural imperative by leveraging visual tactile information to recognize different crops. Participants will be provided with the dataset collected using different fruits and vegetatables. The challenge requires the participants to design innovative solutions utilizing state-of-the-art image processing, computer vision, and machine learning schemes.

Problem-4: Detecting Aquatic Defects Using Underwater Imagery

The health of aquatic environments is critical for ecological balance and human well-being. This conference challenge aims to address this concern by focusing on the recognizing aquatic defects using underwater imagery. Participants will be provided with a specialized dataset comprising underwater images captured in various aquatic environments. The challenge calls for innovative solutions utilizing state-of-the-art image processing, computer vision, and machine learning schemes to recognize different types of aquatic defects.

2. Tournament Phases:

The tournament will be conducted in two phases, where the description of each phase is presented below:

Phase-1: Teams Registration and Problem Selection

In the first phase, the teams will register via Google Forms, and they will select the problem of their choice which they want to solve during the competition. The organizers will also share the datasets privately with the teams with proper legal contract about non re-distribution and non-sharing policies.

Phase-2: Trained Models Submission and Final Evaluations:

In the second phase of the tournament, the participants will submit their codes and trained model weights with the organizers, and the organizers will evaluate their models on the private test dataset. The final evaluation scores and results will then be communicated to the teams. The top-3 winning teams will be given cash prizes during the ceremony. Moreover, all the participating teams will also be given certificates.

3. Datasets and Code Release Policies:

For each problem, the participants will be given a dataset to train their models, the test data will be hidden from the participants for fair evaluation purposes. The participants would be sharing their codes and trained model weights with the organizer who will evaluate the performance of their models. The evaluation scores will be communicated to the participants afterwards.

4. Evaluation Metrics:

The models will be evaluated using the following metrics:

Metrics

5. Prizes and Certificates:

The top-3 teams will be given cash prizes, whereas all the participating teams will be given certificates of appreciation. The distribution of cash prizes w.r.t winning order are given below:

6. Timeline:

7. Eligibility Criteria:

Only students can register for the tournament in the form of teams. A single team can have a maximum of 2 people. The age of the students should be between 20-30 years.

8. Registration:

Registration is CLOSED.

9. Tournament Scoreboard:

The tournament scoreboard is presented below:

Metrics

10. Certificate for Participants:

The certificates can be downloaded from the following link: https://drive.google.com/drive/folders/1OVeG43XW6MyJhKhvWjsJsH0cWgCf-qgq?usp=sharing

11. Important Information:

Please note that the organizers will not be providing any hardware resources for model training, and validation. The teams will need to use their own hardware setup for training the models. If any team lacks hardware resources in their lab/ university, then they are allowed to use free version of Google Colab to train the model and share it with us for evaluation purposes. For model implementation, both PyTorch and MATLAB is allowed.

12. Organizers:

13. Contact Details:

In case of any questions, please feel free to contact us at: taimur.hassan@adu.ac.ae