Anique Akhtar      

Photo 

Anique Akhtar
Senior Engineer at Qualcomm Inc.
Ph.D. Doctorate in Computer and Electrical Engineering from
University of Missouri - Kansas City, Missouri, USA
Advisor: Dr. Zhu Li

Resume (last updated: March 2024)
Google Scholar Link
LinkedIn


Machine learning researcher with experience in deep learning, data compression, computer vision, and generative modeling. Currently working as a Senior Engineer at Qualcomm Inc. working in the Multimedia, Research, and Development (MMRND) team in San Diego, California.

I completed my Ph.D. in Computer and Electrical Engineering from University of Missouri - Kansas City. During my doctorate, I was working under Dr. Zhu Li in the Multimedia Computing & Communication Lab.

I did my M.S. in Electrical Engineering from Koc University, Istanbul, Turkey, where I worked under Dr. Sinem Coleri Ergen in the Wireless Networks Laboratory on 60 GHz directional wireless Communication.

Prior to coming to Koc University, I did my B.Sc in Electrical Engineering at Lahore University of Management Sciences (LUMS), Lahore, Pakistan.

Announcements

  • Qualcomm recognition as an IP contributor. 2023

  • Recipient of the annual award for Outstanding Doctoral Student in Electrical and Computer Engineering for the academic year 2021-22. May 2022

  • Made an official contribution and presented work in MPEG WG 07 3D Graphics Coding meeting regarding dynamic point cloud coding. April 2022

  • Invited to MPEG WG 07 3D Graphics Coding meeting as a point cloud expert. April 2022

  • Dynamic Point Cloud Coding work submitted to a conference. April 2022

  • PU-Dense work accepted in Transactions on Image Processing (TIP). March 2022

  • Dynamic Point Cloud Interpolation work accepted in ICASSP. Feb 2022

  • First inventor on the patent: Point Cloud Geometry Upsampling. U.S. Patent Application 17/345,063. Jan 2022

Research

  • Video-based dynamic mesh coding (V-DMC).

  • MPEG WG7 3DGCH: V-DMC and AI-PCC.

  • Inter-frame and Intra-frame Point Cloud Prediction Schemes.

  • End-to-end Point Cloud Compression.

  • Point Cloud Upsampling.

  • 3D Point Cloud Semantic Segmentation.

  • 3D Point Cloud Denoising and Outlier Removal.

  • Point Cloud Deep Learning Solutions.

  • Low Latency Visual Communication.

  • Neural Networks and Deep Learning.

  • Wireless Networks (LTE and 5G).

  • OFDM and Waveform design.

  • 3GPP RAN.

  • MAC Protocols for Wireless Communication.

  • mmWave Directional Communication.

Project Work

alt text 

Inter-prediction in Point Cloud Compression with Sparse Convolutional Networks

Internship work at Qualcomm.

  • Deep learning solution for inter-frame compression of high resolution point clouds.

  • Proposed I, B, and P-frame encoding framework.

alt text 

PU-Dense: Sparse Tensor-based Point Cloud Geometry Upsampling (Project Page) (GitHub)

Internship work at Qualcomm.

  • A sparse convolution-based point cloud upsampling solution that works on synthetic mesh-based dataset, sparse LiDAR-based point clouds, as well as dense high-resolution photo-realistic point clouds.

  • The current state-of-the-art in point cloud upsampling.

alt text 

Dynamic Point Cloud Interpolation. (pdf)

Internship work at Qualcomm.

  • Deep learning-based point cloud interpolation framework for photorealistic dynamic point clouds.

  • Given two consecutive dynamic point cloud frames, the framework aims to generate intermediate frame(s) between them.

alt text 

Video-based Point Cloud Compression (V-PCC) Artifact Removal. (pdf)

Summer Internship work at Tencent.

  • Video-based Point Cloud Compression (V-PCC) standards introduces quantization at lower bitrate resulting in artifacts in the reconstructed point cloud.

  • We propose a deep learning-based V-PCC artifact removal framework.

  • We exploit the prior knowledge of the direction of quantization noise in V-PCC to learn both the direction as well as level of quantization noise by limiting the degree of freedom of the learned noise.

alt text 

Point Cloud Geometry Prediction Across Spatial Scale. (pdf)

Summer Internship work at Tencent.

  • We propose a deep learning solution for point cloud geometry prediction scheme to upsample a lower Level-of-Detail (LoD) point cloud into a higher LoD point cloud.

  • We employ an octree-type upsampling solution to predict geometry across spatial scale.

alt text 

2D Penoptic Segmentation on street level imagery (SLI) from HERE True Drives.

Summer Internship at HERE Technologies.

  • 2D Building Facade Segmentation and Portal Detection.

  • 2D Building Tracking, Segmentation, and Instant Segmentation.

alt text 

Point Cloud Denoising.

Point clouds obtained from 3D scanners or by image-based reconstruction techniques are often corrupted with a significant amount of non-negligible noise.

  • We propose a two-stage deep neural network that takes in 3D point cloud data and outputs a denoised point cloud.

  • 1st stage: Outlier removal.

  • 2nd stage: Denoising surface noise.

  • We achieve state-of-the-art point cloud denoising results.

(Noisy point cloud on the left, Denoised point cloud on the right. JPEG 8i dataset)

alt text 

3D Semantic Segmentation on HERE True LiDAR Data.

Summer Internship at HERE Technologies.

  • Annotation of large scale outdoor LiDAR point cloud data.

  • Building Deep Learning Architecture for 3D Semantic Segmentation.

  • Feature abstraction from segmented 3D Point Cloud Data.

(Image on the left is from Semantic3D Dataset.)

alt text 

Realtime 3D Point Cloud Communication.

  • Joint source-channel coding for robustness to different channel conditions. (pdf)

  • Adaptive Modulation and Coding schemes for point cloud broadcasting.

  • Adaptive Random Network Coding (ARNC) for scalable point cloud multicasting. (pdf)

  • Low latency support for V2V as well as V2I communication

(Outdoor LiDAR data from Hesai shown on the left.)

alt text 

Mobile Edge Point Cloud Computing.

  • Registering infrastructure based and vehicle based point cloud submaps into one big point cloud.

  • Differentiating the static background from the dynamic live map.

(Google car collecting LiDAR data shown on the left.)

alt text 

Non-Deep Learning-based 3D Point Cloud Geometry Compression.

  • Binary Tree embedded Quad Tree (BTQT) source encoding. (pdf) & (pdf)

  • Lossless point cloud geometry compression.

  • Error Resilient and Scalable point cloud source coding that is layered for different quality of service requirements.

alt text 

Deep Learning (Older Work)

  • Deep Super Resolution Networks for SIFT point repeatability.

    • Novel technique to super-resolve low resolution images into high resolution images while maintaining SIFT key points features.

  • Human Activity Recognition using Wearable Sensors by Deep Convolutional Neural Networks.

    • Designed a Convolutional Neural Network for Human Activity Detection using accelerometers and gyroscopes sensors from your cell-phone or wearable device. The results achieve 95% classification accuracy over 10 classes.

  • Market Trend Prediction for Cryptocurrency using Machine Learning.

    • Using traditional financial Technical Analysis (TA) coupled with deep learning techniques to predict Bitcoin price.

  • Machine learning applications in Wireless Communication and Network Science.

alt text 

5G and Beyond.

  • Resource allocation schemes for 5G heterogenous multi-numerology network. (pdf)

  • Flexible waveform and numerology design for future cellular systems.

  • Adaptive CP size optimization in OFDM waveform.

alt text 

Directional mmWave Communication.

  • Directional MAC Protocol for IEEE 802.11ad WLANs. (pdf) (PROJECT LINK)

  • Efficient Network Level Beamforming Training for IEEE 802.11ad WLANs. (pdf)

  • Energy Efficient MAC Protocol with Localization scheme for Wireless Sensor Networks using Directional Antennas. (pdf draft)

  • Multi-hop network neighbor discovery and beamforming using directional antennas in 802.11ad WLANs. (pdf draft)

  • Optimization of link scheduling in directional wireless networks using Heuristic methods. (pdf draft)

Publications

  • A. Akhtar, Z. Li, and G. Van der Auwera, “Inter-Frame Compression for Dynamic Point Cloud Geometry Coding”
    IEEE Transactions on Image Processing (TIP). 2024. (pdf)

  • A. Akhtar, Z. Li, G. Van der Auwera, L. Li, and J. Chen, “PU-Dense: Sparse Tensor-based Point Cloud Geometry Upsampling” (Project Page) (GitHub)
    IEEE Transactions on Image Processing (TIP). 2022 (pdf)

  • A. Akhtar, Z. Li, G. Van der Auwera, and J. Chen, “Dynamic Point Cloud Interpolation”
    IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). 2022. (pdf)

  • A. Akhtar, W. Gao, L. Li, Z. Li, W. Jia, and S. Liu, “Video-based Point Cloud Compression Artifact Removal”,
    IEEE Transactions on Multimedia (T-MM), 2021. (pdf)

  • W. Jia, L. Li, A. Akhtar, Z. Li, and S. Liu, “Convolutional Neural Network-based Occupancy Map Accuracy Improvement for Video-based Point Cloud Compression”,
    IEEE Transactions on Multimedia (T-MM), 2021.

  • A. Akhtar, W. Gao, L. Li, Z. Li, X. Zhang, and S. Liu, “Point Cloud Geometry Prediction Across Spatial Scale using Deep Learning”,
    IEEE Visual Communication & Image Processing Conf (VCIP), Hong Kong, 2020. (pdf)

  • A. Akhtar, J. Ma, R. Shafin, J. Bai, L. Li, Z. Li, L. Liu, “Low Latency Scalable Point Cloud Communication in VANETs using V2I Communication”
    IEEE International Conference on Communications (ICC), Shanghai, China. 2019. (pdf)

  • A. Akhtar, B. Kathariya, Z. Li, “Low Latency Scalable Point Cloud Communication”
    IEEE International Conference on Image Processing (ICIP), Taipei, Taiwan. 2019. (pdf)

  • A. Akhtar, H. Arslan, “Downlink Resource Allocation and Packet Scheduling in Multi-Numerology Wireless Systems”
    IEEE Wireless Communications and Networking Conference (IEEE WCNC), 2018. (pdf)

  • A. Akhtar, S. Coleri Ergen, “Directional MAC Protocol for IEEE 802.11ad WLANs”
    Ad Hoc Networks, 2018. (pdf) (Protocol's website with explanation and open source code)

  • A. Akhtar, S. Coleri Ergen, “Efficient Network Level Beamforming Training for IEEE 802.11ad WLANs”
    International Symposium on Performance Evaluation of Computer and Telecommunication Systems (SPECTS 2015), Chicago, Illinois, US. July 2015 (pdf)

Patents

  • A. Akhtar, W. Gao, X. Zhang, and S. Liu, Tencent America LLC, 2024. Point cloud geometry upsampling. U.S. Patent 11,893,691.

  • A. Akhtar, G. Van der Auwera, A.K. Ramasubramonian, M. Karczewicz and L.P. Van, Qualcomm Inc, 2023. Geometry coordinate scaling for AI-based dynamic point cloud coding. U.S. Patent Application 18/318,498.

Noticeable MPEG Contributions

  • Anique Akhtar, Geert Van der Auwera, Reetu Hooda, Adarsh Krishnan Ramasubramonian, Marta Karczewicz, “[V-DMC][New] Implementation of Normal Encoding in V-DMC TMM v6.0 ”, MPEG-145 Online, Doc. m66553, Jan 2024.

  • Anique Akhtar, Geert Van der Auwera, Reetu Hooda, Adarsh Krishnan Ramasubramonian, Marta Karczewicz, “[V-DMC][New] On Basemesh”, MPEG-144 Hannover, Doc. m65333, Oct 2023.

  • Anique Akhtar, Geert Van der Auwera, Adarsh Krishnan Ramasubramonian, Marta Karczewicz, “[AI-3DGC][EE 5.5 related] Software release for Hybrid AI-based Geometry + G-PCC Attribute Coding”, MPEG-143 Geneva, Doc. m64441, July 2023.

  • Anique Akhtar, Geert Van der Auwera, Birendra Kathariya, Adarsh Krishnan Ramasubramonian, Marta Karczewicz, “[AI-3DGC][EE5.3-related] Update on baseline attribute compression for ML-based PCC”, MPEG-142 Antalya, Doc. m63255, April 2023.

  • Anique Akhtar, Geert Van der Auwera, Birendra Kathariya, Adarsh Krishnan Ramasubramonian, Marta Karczewicz, “[AI-3DGC][EE5.3-related] Baseline attribute compression for ML-based PCC”, MPEG-140 Mainz, Doc. m61313, Oct 2022.

  • Anique Akhtar, Geert Van der Auwera, Adarsh Krishnan Ramasubramonian, Marta Karczewicz, “[AI-3DGC][EE5.3 Test 2] Results dynamic point cloud compression”, MPEG-140 Mainz, Doc. m61201, Oct 2022.

  • Anique Akhtar, Zhu Li, Geert Van der Auwera, Adarsh Krishnan Ramasubramonian, Luong Pham Van, Marta Karczewicz, “Dynamic Point Cloud Geometry Compression using Sparse Convolutions”, MPEG-137 Online, Doc. m59617, April 2022.

Work Experience

  • Qualcomm. Current: (January. 2022 - Present)

    • Video-based dynamic mesh coding (V-DMC).

    • MPEG WG7 3DGCH: V-DMC and AI-PCC.

    • Inter-Frame and Intra-Frame Dynamic Point Cloud Compression using Deep Learning.

    • Deep Learning-based Point Cloud Interpolation.

    • Point Cloud Upsampling using Deep Learning.

  • Tencent. 3 months: (June. 2020 - Aug 2020)

    • Video-based Point Cloud Compression (V-PCC) Artifact Removal.

    • Point Cloud Geometry Prediction.

    • 3D Point Cloud Denoising.

    • End-to-End Point Cloud Compression (PCC).

  • HERE Technologies. 6 months: (June. 2019 - Nov 2019)

    • 2D Building Tracking, Segmentation, and Instance Segmentation.

    • 2D Facade Segmentation and Portal Detection.

    • 3D Point Cloud Semantic Segmentation.

Unpublished Work

  • A. Akhtar, Z. Li, “Point Cloud Denoising using Deep Neural Networks”
    2020. Outlier removal as well as surface denoising to achieve state-of-the-art denoising on point cloud data.

  • A. Akhtar, Y. Yilmaz, “Machine Learning for Market Trend Prediction in Bitcoin”
    2017. Using traditional financial Technical Analysis (TA) coupled with deep learning techniques to predict Bitcoin price. (pdf draft)

  • A. Akhtar, S. Coleri Ergen, “Energy Efficient MAC Protocol with Localization scheme for Wireless Sensor Networks using Directional Antennas.”
    2015. Energy efficient MAC protocol for with localization scheme using beamforming. (pdf draft)

  • A. Akhtar, O. Ozkasap, “Multi-hop network neighbor discovery and beamforming using directional antennas in 802.11ad WLANs”
    2014. Unpublished. (pdf draft)

  • A. Akhtar, Ceyda Oguz, “Optimization of link scheduling in directional wireless networks using Heuristic methods”
    2014. We propose multiple heuristic methods to solve the problem at hand and discuss how the problem behaves in different scenarios. (pdf draft)

Education

  • Ph.D. Computer and Electrical Engineering. - August 2016 - present

    • University of Missouri - Kansas City, Missouri, USA. - January 2018 - present
      Expected Graduation: January 2021

      • Advisor: Dr. Zhu Li

      • Research: Multimedia Computing & Communication.

    • University of South Florida, Tampa, Florida, USA. - August 2016 - December 2017

      • Research: 5G and Beyond, Machine Learning & Data Science.

  • Master of Science, Electrical Engineering - 2013-2015
    Koc University, Istanbul. Turkey
    Graduated: July 2015

  • Bachelor of Science, Electrical Engineering - 2008-2013
    Lahore University of Management Science, Lahore, Pakistan
    Graduated: June 2013

Teaching Experience

  • University of Missouri-Kansas City January. 2018 - Present
    Department of Computer Science

    • ECE/CS 479/5582, Computer Vision - Fall 2021

    • ECE/CS 5578, Multimedia Communication - Spring 2021

    • ECE/CS 484, Digital Image Processing - Fall 2020

  • Koc University Aug. 2013 - July 2016
    Department of Electrical Engineering

    • ELEC 317, Microprocessors - Fall 2013

    • ENG 200, Probability for Engineers - Spring 2014

    • SCI 100, Natural Sciences - Fall 2014

    • ENG 200, Probability for Engineers - Spring 2015

    • SCI 100, Natural Sciences - Fall 2015

    • ENG 200, Probability for Engineers - Spring 2016

  • Lahore University of Management Science Aug. 2012- Jan. 2013
    Department of Electrical Engineering

    • EE 421, Digital System Design - Fall 2012