I am currently seeking opportunities in Berlin, Germany, focusing on the development of Large Language Models and advanced AI systems. My extensive experience in R&D for computer vision and multimodal applications positions me to effectively lead and innovate in this field.
- Technical Lead in R&D for Computer Vision
- End-to-end generative AI pipeline for cloud-based solutions
- Multimodal LLMs for image processing applications
- Large Vision Models (LVMs) for edge computing
- Engineered end-to-end, cloud-based generative AI solutions, overseeing the entire pipeline from data ingestion and model training to deployment and scaling.
- Expertise in Multimodal Large Language Models (LLMs):
- Specialized in integrating multimodal LLMs for image processing applications, enhancing system capabilities to interpret and analyze visual and textual data simultaneously.
- Innovations with Large Vision Models (LVMs):
- Developed and optimized Large Vision Models for edge computing, ensuring efficient processing and responsiveness in IoT devices.
- Multimodal RAG Systems: Led the development of Retriever-Augmented Generation (RAG) applications integrating text, image, and structured data, enhancing multimodal interaction capabilities.
- Advanced AI Pipelines: Engineered end-to-end solutions for generative AI, leveraging cloud-based architectures to deploy scalable and efficient AI systems.
- Deep Learning Implementation: Proficient in implementing complex deep learning models, with extensive use of libraries such as PyTorch, OpenAI's GPT models, and langchain for sophisticated text and image processing tasks.
- Data Handling and Processing: Experienced in manipulating large-scale datasets, implementing custom extraction and partition techniques for PDF data integration, utilizing Python's robust libraries like PyPDF2 and pytesseract for OCR functionalities.
- Optimization Techniques: Applied advanced machine learning techniques including hyper-parameter tuning, quantization, and model compression to enhance performance and efficiency on target hardware platforms, particularly in edge computing scenarios.
- AI Model Deployment: Skilled in deploying AI models using Docker, managing environments with dependencies including langchain, unstructured, PyPDF2, and various OpenAI services, ensuring smooth transition from development to production.
- Research and Development: Authored comprehensive documentation and guides, effectively summarizing research findings and technical processes, demonstrated through detailed GitHub repositories and Jupyter notebooks.
- AI-Powered Summarization: Developed capabilities for summarizing diverse data elements (text, tables, images) using AI-driven approaches, significantly improving information accessibility and user engagement.
- Community Contribution and Collaboration: Actively engaged in community forums and collaborative projects, contributing to open-source projects and providing innovative solutions to complex problems in the AI space.
I am eager to bring my expertise to a dynamic team in Berlin, where I can contribute to groundbreaking projects and further advance the field of artificial intelligence.
- Patents: A METHOD FOR AUGMENTING A PLURALITY OF FACE IMAGES - 2021
- The present invention relates to a method for increasing data for face analysis in video surveillance.
- WO2021060971A1
- Patents: A METHOD FOR DETECTING A MOVING VEHICLE - 2021
- The present invention relates to a method for detecting a moving vehicle.
- WO2021107761
- Patents: System and method for providing advertisement contents based on facial analysis - 2020
- Invented an algorithm, methods, and system for advanced facial attribute detection, leading to improvements in advertising systems.
- WO2020141969A2 WIPO (PCT)
- Book Chapter: Camera Calibration and Video Stabilization for Robot Localization, Springer, 2021.
- Authored over 16 publications in books, journals, and conferences globally.
- Image Processing GPT
- GPTs: Computer Vision Developer
- Expert in Python, OpenCV for image processing and computer vision applications.
- MindMap about LLMs & LLMOps
- Code for chat app with OpenRouter's AI! 🚀 Utilize asyncio and aiohttp for seamless conversations and manage interactions with a smart queue. Dive into the future of chat applications now!"
- fine-tune LLMs
- Microsoft AI Lab: RAG Workflow with Azure AI
- Lab Focus: Hands-on RAG workflow development using Azure AI Studio and Prompt Flow.
- Skills Acquired: Mastery in LLMOps, Azure AI Studio usage, and Prompt Flow integration.
- Tools Used: GitHub Codespaces, Visual Studio Code, Azure AI & ML Studio, Azure Portal.
- Outcome: Successfully developed and deployed "Contoso Chat", enhancing skills in scalable AI solution development.
https://github.com/pirahansiah/cvtest
The first function is int func_image_info(cv::Mat src, cv::Mat &dst /*output*/) this function show information about image such as size, histogram, ....
YouTube link for OpenCV: https://www.youtube.com/watch?v=gK1ybsWOqhs
I have 6+ years of experience as a computer vision research engineer in three multinational companies in two continents, strengthened by my academic background with a Master’s and PhD in Computer Science (Computer Vision). My expertise includes Technical Lead R&D, Software Specialists Image Processing - Medical Devices, Computer Vision with Machine Learning (Object Detection, Video Tracking), IoT, and Robotics; and I am experienced in designing algorithms for Image Thresholding, Optical Flow, Camera Calibration, and Stereo Vision. Lastly, I have a track record in creating effective metrics, building end-to-end pipelines, and writing production-level codes with OpenCV and Deep Learning frameworks (Caffe, TensorFlow, PyTorch).
FarshidPirahanSiah
I am interested in Metaverse, Medicine. I am interested in 3D Camera Calibration for extended reality headset in Metaverse.
I have experience in computer vision, deep learning and robotic.
I am familiar with IoT and Edge computing, Medical devices, cloud base solution (AWS), robotic.
Platform for metaverse
AR/VR Frameworks Engineer For New Application Paradigm
Geometric camera calibration, also referred to as camera re-sectioning, estimates the parameters of a lens and image sensor of an image or video camera. These parameters can be used to correct for lens distortion, measure the size of an object in world units, or determine the location of the camera in a scene. These tasks are used in applications such as machine vision to detect and measure objects. They are also used in robotics, navigation systems, and 3-D scene reconstruction. Without any knowledge of the calibration of the cameras, it is impossible to do better than projective reconstruction (MathWorks).
Non-intrusive scene measurement tasks, such as 3D reconstruction, object inspection, target or self-localization or scene mapping require a calibrated camera model (Orghidan et al. 2011). Camera calibration is the process of approximating the parameters of a pinhole camera model (Tsai 1987; Stein 1995; Heikkila & Silven 1997) of a given photograph or video.
Camera self-calibration, also known as auto/fully calibration method, is not reliant upon the calibration reference object of a camera. Three-dimensional reconstruction and motion estimation are two fundamental tasks in computer vision (Kaehler & Bradski 2016). In both tasks, camera calibration is an essential step that bridges the 2D imaging plane and 3D space. For the past decade, camera calibration has been heavily investigated in the fields of computer vision and optics (Anuar et al. 2015; Garg & Deep 2015; Hong et al. 2015; Jia et al. 2015). Maybank and Faugeras (1992) introduced the concept of camera self-calibration. However, the self-calibration method is nonlinear and highly sensitive to noises; these methods can be enhanced by using active vision, where some specific camera motions are designed, such as pure rotation, orthogonal translations (Wang et al. 2004). For example, Hartley proposed using pure rotation to compute the infinite homography, then linearly calibrate the camera (Hartley & Zisserman 2003). However, the constraints on the specific motions are too strong to satisfy in practice, which hinder them from wider applications (Lei et al. 2004). For example, it is difficult to perform pure rotation around the camera’s optical center, even with a pure rotation platform, because it is difficult to obtain the camera’s optical center and even more difficult to coincide the camera’s optical center with the rotation center of a rotation platform. Furthermore, some researchers tried to improve self-calibration using more constraints, such as module constraint and loop constraint (Courchay et al. 2012). Another category of calibration methods is usually based on specific calibration rig or scene constraints (Liming et al. 2013).
The first step for camera calibration is corner detection. Based on my research, the calibration pattern image play important rule in the whole calibration process.
-
Camera calibration for multi-modal robot vision based on image quality assessment https://www.researchgate.net/profile/Farshid-Pirahansiah/publication/288174690_Camera_calibration_for_multi-modal_robot_vision_based_on_image_quality_assessment/links/5735bc2908aea45ee83c999e/Camera-calibration-for-multi-modal-robot-vision-based-on-image-quality-assessment.pdf
-
Pattern image significance for camera calibration https://ieeexplore.ieee.org/abstract/document/8305440
-
Camera Calibration and Video Stabilization Framework for Robot Localization https://link.springer.com/chapter/10.1007/978-3-030-74540-0_12
- CV_metaverse
- 3D_multi_camera_calibration
- corner_Detection
- cornerDetection.ipynb
- auto multi camera calibration
- corner_Detection
- 3D_multi_camera_calibration
Top source code:
- cornerDetection.ipynb
- It use several preprocessing and postprocessing steps to enhance corner detection use by camera calibration.
- 3D multi camera calibration require detect and set points for all camera together
- if the calibration pattern images are not good, blur, ... it need to enhance it first then use corner points to detect and use for calibration process
#computervision #AI #objectdetection #objecttracking #ml #research #CNN #gans #convolutionalneuralnetworks #ai #vr #reinforcementlearning #mlops #aiforbusiness #science #researcher #phd #cameracalibration #opticalflow #videostablization #humanoidrobot #localization #3dSLAM #reconstruction #pointcloud #mixedreality #edgecomputing #raspberrypi #intelstick #googlecoral #jetsonnano #nvidiavgpu #tensorflowjs #pytorch #opencv #aikit #caffee #DIGITS #c++ #python #ubuntu #farshidpirahansiah #tiziran.com #farshid #pirahansiah #robotics #tiziran.com #farshid #pirahansiah #MultiCameraMultiClassMultiObjectTracking #deeplearning #machinelearning #artificialintelligence #tensorflow #robotics #3dvision #sterovision #depthmap #RCNN #machinevision #imageprocessing #patternrecognition #compiler #RISC-V #RNN #fullStackDeepLearning #productinnovation #patents #TensorRT #ApacheTVM #TFLite #PyTorchmobile #dockers #gRPC #RESTAPIs #GRPC #GraphQL #imageprocessing #patternrecognition #EnablingEfficient #high-performance #Accelerators #Optimization #computervision #AI #objectdetection #objecttracking #ml #research #CNN #gans #convolutionalneuralnetworks #ai #vr #reinforcementlearning #mlops #aiforbusiness #science #researcher #phd #cameracalibration #opticalflow #videostablization #humanoidrobot #localization #3dSLAM #reconstruction #pointcloud #AR/VR #mixedreality #edgecomputing #raspberrypi #intelstick #googlecoral #jetsonnano #nvidiavgpu #tensorflowjs #pytorch #opencv #aikit #caffee #DIGITS #c++ #python #ubuntu #farshidpirahansiah #tiziran.com #farshid #pirahansiah #robotics #SingleObjecttracking #SOT #MultiObjecttracking #MOT #MultiTargetTracking #MTT #MultiClassMultiObjecttracking #MCMOT #MultiCameraMultiClassMultiObjectTracking #MCMCMOT #deeplearning #machinelearning #artificialintelligence #computervision #video #objectdetection #objecttracking #tensorflow #innovation #learning #datascience #robotics #3dvision #sterovision #depthmap #RCNN #machinevision #imageprocessing #patternrecognition #compiler #RISC-V #RNN #fullStackDeepLearning #productinnovation #patents #TensorRT #ApacheTVM #TFLite #PyTorchmobile #TensorFlow.js #CoreML #MLkit #DataDog #NewRelic #AmazonCloudWatch #dockers #gRPC #RESTAPIs #GRPC #GraphQL #farshidpirahansiah #tiziran.com #farshid #pirahansiah #robotics #SingleObjecttracking #SOT #MultiObjecttracking #MOT #MultiTargetTracking #MTT #MultiClassMultiObjecttracking #MCMOT #MultiCameraMultiClassMultiObjectTracking #MCMCMOT #deeplearning #machinelearning #artificialintelligence #computervision #video #objectdetection #objecttracking #tensorflow #innovation #learning #datascience #robotics #3dvision #sterovision #depthmap #RCNN #objectdetection #objecttracking #ml #research #CNN #gans #convolutionalneuralnetworks #ai #vr #reinforcementlearning #mlops #aiforbusiness #science #researcher #phd #cameracalibration #opticalflow #videostablization #humanoidrobot #localization #3dSLAM #reconstruction #pointcloud #AR/VR #mixedreality #edgecomputing #raspberrypi #intelstick #googlecoral #jetsonnano #nvidiavgpu #tensorflowjs #pytorch #opencv #aikit #caffee #DIGITS #c++ #python #ubuntu #machinevision #imageprocessing #patternrecognition #SingleObjecttracking #SOT #MultiObjecttracking #MOT #MultiTargetTracking #MTT #MultiClassMultiObjecttracking #MCMOT #MultiCameraMultiClassMultiObjectTracking #MCMCMOT #deeplearning #machinelearning #artificialintelligence #computervision #video #objectdetection #objecttracking #tensorflow #innovation #learning #datascience #robotics #3dvision #sterovision #depthmap #RCNN #objectdetection #objecttracking #ml #research #CNN #GAN
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent