Venu Madhav Nookala

Venu Madhav Nookala

Auto-generated profile, based on publicly available information

Venu Madhav Nookala

KPIT Technologies, Bengaluru, India

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
26 September 2023 in IEEE Access
Verified
Journal Article
Open access
RSRaghavendra Somappa
Raghavendra Somappa
SAS. K. Abhilash
S. K. Abhilash
VNVenu Madhav Nookala
Venu Madhav Nookala
SNSrinidhi N N
Srinidhi N N
NAN. D. Adesh
N. D. Adesh
Over the past few years, deep learning techniques have revolutionized the field of face parsing by utilizing massive datasets to generate high-level features and achieve outstanding performance. Usually, these techniques involve Convolutional Neural Networks (CNNs) to derive features from the input image and then a decoder network to forecast the semantic labels for every pixel. Even with complex deep convolutional neural networks (DCNN), incorrect parsing leads to a semantic gap between identical features, especially at boundary levels. Face parsing has encountered additional challenges due to the inclusion of pose, illumination and facial expressions. In order to address these issues, a Boundary Aware Dilated Attention Network (BADANet) is introduced which explores the use of multi-scale techniques to improve the accuracy and robustness of the per-pixel frames. BADANet’s dilated attention module has been combined with a lightweight backbone to achieve exceptional results on LaPa, CelebAMask-HQ, and iBugMask. Extensive evaluations of the proposed method demonstrate its performance to be on par with various state-of-the-art techniques. The pipeline for the BADANet is available at https://github.com/abhigoku10/BADANet.

26 June 2025 in IEEE Access
Verified
Journal Article
Open access
ANAnoop B N
Anoop B N
SAS. K. Abhilash
S. K. Abhilash
VNVenu Madhav Nookala
Venu Madhav Nookala
SPS. Prita
S. Prita
RSRaghavendra Somappa
Raghavendra Somappa
Road traffic accidents claim 1.35 million lives annually, making them the leading cause of death among individuals aged 5–29 years, according to the world health organization(WHO). Low- and middle-income countries, despite having only 60% of the world’s vehicles, account for 93% of fatalities due to speeding, impaired driving, inadequate safety measures, and poor infrastructure. In the era of Industry 4.0, intelligent driving assistance systems offer a transformative solution to mitigate human errors, particularly those caused by fatigue or drowsiness. This study presents You Look Once for Multi-Feature(YLOMF), a novel single-stage, vision-based hierarchical model for autonomous driving. It features modular feature component heads, ensuring customizable deployment and real-time efficiency on edge devices. Trained on the Karlsruhe Institute of Technology and Toyota Technological Institute (KITTI) dataset and evaluated on Berkeley Deep Drive Dataset (BDD100K), Indian Driving Dataset (IDD), and Audi Autonomous Driving Dataset (A2D2), YLOMF exhibits robust cross-domain generalization. It achieves 85.1% IoU for lane segmentation, 81.56% accuracy in 2D object detection, and 31.5% (easy), 26.38% (medium), and 23.9% (hard) accuracy in 3D instance detection, surpassing state-of-the-art benchmarks. Depth estimation performance was validated using absolute relative error, squared relative error, RMSE, and logged RMSE. By delivering a computationally efficient and highly accurate perception framework, YLOMF enhances scene understanding and object recognition in real time. Its integration into autonomous systems offers significant potential for reducing road accidents and improving overall traffic safety in safety-critical environments.

21 May 2025 in IEEE Access
Verified
Journal Article
Open access
RSRaghavendra Somappa
Raghavendra Somappa
RKRajalekshmi Kishore
Rajalekshmi Kishore
SAS. K. Abhilash
S. K. Abhilash
VNVenu Madhav Nookala
Venu Madhav Nookala
Manish Varun YadavManish Varun Yadav
Manish Varun Yadav
The need for real-time autonomous driving reveals significant gaps in terms of accuracy rates and processing efficiency. Current systems struggle to maintain high accuracy across all scenarios while ensuring efficient real-time processing. Addressing these gaps, the proposed work presents a novel contribution aimed at advancing the capabilities of autonomous driving systems. The model introduces a single network architecture capable of simultaneously handling multiple crucial tasks including dynamic and static segmentation, depth estimation, and 3D object detection. The proposed architecture is specifically designed to be lightweight, ensuring efficient deployment for real-time autonomous driving applications. Moreover, the proposed single-stage architecture is end-to-end trainable, offering a streamlined approach to model training. Through extensive experimentation, the results demonstrate superior performance over existing architecture, achieving remarkable accuracy rates. Overall, the proposed work represents a significant step forward in the development of efficient and effective solutions for real-time autonomous driving.

3 August 2023 in IEEE Access
Verified
Journal Article
Open access
RSRaghavendra Somappa
Raghavendra Somappa
DRDivya Rao
Divya Rao
SAS. K. Abhilash
S. K. Abhilash
VNVenu Madhav Nookala
Venu Madhav Nookala
PBPraveen Gurunath Bharathi
Praveen Gurunath Bharathi
Amodal segmentation is a critical task in the field of computer vision as it involves accurately estimating object boundaries that extend beyond occlusion. This paper introduces a network named after the Amodal Segmentation Head, ASH-Net, a novel architecture specifically designed for amodal segmentation. ASH-Net is comprised of a ResNet-50 backbone, a Feature Pyramid Network middle layer, and an Amodal Segmentation Head. The evaluation encompasses three diverse datasets, namely COCOA-cls, KINS, and D2SA, providing a comprehensive analysis of ASH-Net’s capabilities. The results obtained demonstrate the superiority of ASH-Net in accurately estimating object boundaries beyond occlusion across multiple datasets. Specifically, ASH-Net achieves an Average Precision of 62.15% on the COCOA-cls dataset, 72.58% on the KINS dataset, and an impressive 91.4% on the D2SA dataset. Through extensive evaluation using average precision and average recall metrics, ASH-Net exhibits exceptional performance compared to state-of-the-art models. These findings highlight the remarkable performance of ASH-Net in overcoming occlusion challenges and accurately delineating object boundaries. This research identifies optimal training parameters, such as coefficient dimensions and aspect ratios, that significantly enhance segmentation performance while maintaining computational efficiency. The proposed ASH-Net architecture and its performance pave the way for improved object recognition, enhanced scene understanding, and the development of practical applications in various domains.

14 July 2025 in IEEE Access
Verified
Journal Article
Open access
RSRaghavendra Somappa
Raghavendra Somappa
VIVijayalakshmi
Vijayalakshmi
VAVainidhi
Vainidhi
SAS. K. Abhilash
S. K. Abhilash
VNVenu Madhav Nookala
Venu Madhav Nookala
RARamyashree
Ramyashree
The generation of high-quality annotated image datasets with low computational cost and automated labeling is essential for advancing computer vision systems. However, manual labeling of real images is often labor intensive and expensive. To overcome these challenges, proposed a model named SVPDSA, a generic dataset generation model that incorporates residual and attention block pruning, reduced sampling steps, and network quantization. The model comprises two main training phases: Refined LDM((Latent Diffusion Model) retraining and P-Decoder(Perception decoder) training. A compressed Unet-based diffusion model, pre-trained on the LAION-5B dataset, serves as the foundation for efficient text-to-image synthesis. The model is trained for approximately 50,000 iterations with a learning rate of 0.0001, ensuring lightweight yet effective generation. The proposed model efficiently generates diverse synthetic images with high-quality perception annotations. The proposed approach utilizes a lightweight trained diffusion model and extends text-guided image synthesis to perception data generation, ensuring the quality of the generated datasets while offering a flexible solution for label generation. A decoder module is introduced to expand latent code features and generate labeled annotations for tasks such as semantic segmentation, instance segmentation, and depth estimation. Training the decoder requires fewer than 100 manually labeled images, enabling the creation of an infinitely large annotated dataset. Evaluation on the Cityscapes dataset demonstrates that SVPDSA matches or surpasses existing methods like Mask2Former and DatasetDM in key object classes, including cars, buses, and bicycles. It achieves a mean IoU of 42.7 with ResNet-50 and 41.4 with Swin-B using only 9 real images and 38k synthetic samples, showcasing its efficiency in generating high-quality annotations with minimal real data. Deploying the proposed models on edge devices results in less than a 5-second inference time.This research contributes toward building resource-efficient data generation systems suitable for constrained training environments.

21 July 2025 in IEEE Access
Verified
Journal Article
Open access
RARamyashree
Ramyashree
RSRaghavendra Somappa
Raghavendra Somappa
SAS K Abhilash
S K Abhilash
VNVenu Madhav Nookala
Venu Madhav Nookala
AKArun Kumar
Arun Kumar
PMP Malashree
P Malashree
Generating high-fidelity surround view images from text prompts is a complex task that requires balancing contextual coherence with computational efficiency. The proposed work introduces a novel methodology that combines recurrent attention-based encoder-decoder architectures with text-to-image diffusion models to produce coherent and continuous surround view images. The approach utilizes a custom text encoder to convert input text prompts into contextual embeddings, which are then processed by the proposed ViewNet Unet2d architecture within the decoder. This architecture employs dual cross-attention mechanisms: one aligns text embeddings with corresponding noise image latents, while the other integrates previously generated image latents to ensure continuity across the sequence. This method guarantees that each generated image adheres to its specific prompt, while maintaining coherence with preceding images. In addition, an annotation decoder was introduced that generates semantic segmentation maps, instance segmentation masks, and object detection annotations. The annotation decoder processes latent image maps using a shared feature extraction backbone and dedicated heads for each annotation task.Experimental results on the nuScenes validation set demonstrate the effectiveness of the proposed model in producing high-quality contextually aligned surround view images. The proposed model achieves an FVD of 99 and an FID of 12.6, outperforming existing methods such as Panacea+ and DriveDreamer-2. Furthermore, our approach improves segmentation and detection accuracy, achieving a PQ of 67.4, mIoU of 80.1, and mAP of 65.4, surpassing methods like OpenSeeD and D2Det. An ablation study highlights the contributions of key components in our architecture. Integrating positional encoding, self-attention, and concurrent attention significantly enhances generation quality, reducing FVD to 99 and FID to 12.6. Experimental results demonstrate the effectiveness of proposed work in producing high-quality, contextually aligned surround view images with comprehensive annotations, pushing the boundaries of text-to-image synthesis and scene understanding.