Open Project
Isolated Sign Language Recognition [Google - Isolated Sign Language Recognition | Kaggle ]
Around 90% of which are born to hearing parents many of which may not know American Sign Language. (kdhe.ks.gov, deafchildren.org) Without sign language, deaf babies are at risk of Language Deprivation Syndrome. This syndrome is characterized by a lack of access to naturally occurring language acquisition during their critical language-learning years. It can cause serious impacts on different aspects of their lives, such as relationships, education, and employment. The goal of this competition is to classify isolated American Sign Language (ASL) signs.
Dataset: Google - Isolated Sign Language Recognition | Kaggle
Mentor: Mrugendrasinh Rahevar
Project Status: Open
Team-size: 03
Technology: Python and Pytorch or Keras
Aerial View or Sports Action Recognition using Deep Learning
Human activity recognition from unconstrained videos is still very challenging due to large variations in human body pose, differences in the appearance of interacted objects, occlusions, and inter and intraclass variations. This can be even more challenging for aerial view videos due to additional challenges such as diminutive size, camera motion, significant camera rotation with respect to the target, out of view movement. Prior studies demonstrate the influence of robust action recognition systems focused on ground-level videos. However, the performance of such systems on aerial view videos is questionable.
Dataset: Okutama-Action | Okutama-Action: An Aerial View Video Dataset for Concurrent Human Action Detection
Mentor: Mrugendrasinh Rahevar
Project Status: Open
Team-size: 03
Technology: Python and Pytorch or Keras
CT-Cervical spine fracture detection
Bone fracture is increasing rapidly nowadays due to traumatic injuries and this may also lead to death. so, it becomes important to detect the exact location of the fracture as soon as possible this can help the radiologist to find the location of the fracture so they can treat the patient as soon as possible. There cannot be a generalized method for the detection of bone fracture in humans as every body part of a human being consist of different shapes and sizes of bones and there is less research done on cervical spine fracture detection. In this, we will be using CT scan images in DICOM format and apply Attention on we need to find a fracture that occurred only cervical spine (C1-C7). In the medical field up till now only CNN has applied its extensive models. So, we are planning to apply the attention model on it.
Dataset: RSNA 2022 Cervical Spine Fracture Detection | Kaggle
Mentor: Mrugendrasinh Rahevar
Technology: Python Pytorch/Keras
Project Status:
Team-Size: 02 (6th Semester) (02 8th Semester)
Technology: Python and Pytorch or Keras
Skeleton Based Action recognition using Graph Neural Network
The Human body skeleton acted as a spatiotemporal graph, which is the key inspiration for researchers to implement a GCN-based method for action recognition. Most recently proposed GCN-based techniques combine convolutions with a self-attention mechanism to extract the most informative joint of a human skeleton and increase the model accuracy. The traditional multi-head attention method absorbed all attention heads equally, whereas in most situations, especially for extended skeleton sequences, not all attention heads are expected
Dataset: ROSE Lab (ntu.edu.sg)
Mentor: Mrugendrasinh Rahevar
Technology: Python, Pytorch/Keras
Team-Size: 04
Small Object Detection Challenge for Spotting Birds
The task of this challenge is to detect relatively small birds with few pixel features on images, which is categorized to the Small Object Detection (SOD) problem. Distant-bird detection is an important function for unmanned aerial vehicles (UAVs) such as drones to avoid bird attacks, or drive away harmful birds that destroy fields and rice paddies. Thus, this challenge not only raises academic issues in Computer Vision but also promotes practical technology developments that are expected to be employed in real-world applications. The images in our dataset are captured from drones. In addition to the general difficulties of SOD (e.g., lack of geometrical information for small objects), the annotated birds are of different types (e.g., hawk, crow, and wild bird), and have different parallax, postures, and different degrees of motion blur.
Dataset: Small Object Detection Challenge for Spotting Birds 2023 - MVA2023 (mva-org.jp)
Mentor: Mrugendrasinh Rahevar
Technology: Python, Pytorch/Keras
Team-Size: 03
Predicting 6 Vital Plant Traits from Plant Images for Ecosystem Health
Think of plants as the superheroes of our ecosystems. Their traits hold the key to understanding ecosystems, e.g., in terms of their diversity, productivity, or how these green heroes face the challenges brought on by climate change. By solving this competition effectively, you become a superhero, too — contributing to our understanding of how plants navigate the complexities of climate change. Our goal is to predict a broad set of 6 plant traits (e.g. leaf area, plant height) from crowd-sourced plant images and some ancillary data.
Dataset: PlantTraits2024 - FGVC11 | Kaggle
Mentor: Mrugendrasinh Rahevar
Technology: Python, Pytorch/Keras
Team-Size: 03