You will get Computer Vision: Object Detection, Face Recognition & Anti-Spoofing Systems
Project details
I build computer vision systems that go beyond basic detection, verification, liveness, and behavior analysis included. Recent work spans:
Object detection & classification: trained YOLOv8 models combined with CLIP embeddings and clustering to classify objects with fine-grained categories (e.g. distinguishing near-identical variants that basic detection alone can't tell apart).
Face verification & liveness/anti-spoofing: built a multi-stage pipeline (YOLOv8 → RetinaFace → DeepFace) for real-time identity verification, plus a separate anti-spoofing system combining phone-in-frame detection, screen-artifact detection, and flash-response liveness checks, built to catch photo/screen spoofing attempts on mobile devices.
Pose estimation & activity recognition: multi-backend human pose pipelines with a custom-trained LSTM classifier for activity detection, including real-time form-checking using camera-angle-robust metrics.
I work fast using modern AI tooling without cutting corners on model quality or edge-case handling, spoofing detection and false-positive/negative tradeoffs get real attention, not an afterthought.
Object detection & classification: trained YOLOv8 models combined with CLIP embeddings and clustering to classify objects with fine-grained categories (e.g. distinguishing near-identical variants that basic detection alone can't tell apart).
Face verification & liveness/anti-spoofing: built a multi-stage pipeline (YOLOv8 → RetinaFace → DeepFace) for real-time identity verification, plus a separate anti-spoofing system combining phone-in-frame detection, screen-artifact detection, and flash-response liveness checks, built to catch photo/screen spoofing attempts on mobile devices.
Pose estimation & activity recognition: multi-backend human pose pipelines with a custom-trained LSTM classifier for activity detection, including real-time form-checking using camera-angle-robust metrics.
I work fast using modern AI tooling without cutting corners on model quality or edge-case handling, spoofing detection and false-positive/negative tradeoffs get real attention, not an afterthought.
Machine Learning Tools
GitHub Copilot, Keras, NumPy, OpenCV, pandas, Python, Python Scikit-Learn, PyTorch, TensorFlowWhat's included
| Service Tiers |
Starter
$149.90
|
Standard
$399.90
|
Advanced
$799.90
|
|---|---|---|---|
| Delivery Time | 5 days | 10 days | 16 days |
Number of Revisions | 1 | 2 | 3 |
Number of Model Variations | 1 | 1 | 1 |
Number of Scenarios | 3 | 5 | 8 |
Number of Graphs/Charts | 1 | 2 | 3 |
Model Validation/Testing | |||
Model Documentation | - | ||
Data Source Connectivity | - | - | |
Source Code | - |
Optional add-ons
You can add these on the next page.
Fast Delivery
+$30 - $70
Additional Revision
+$30
Additional Model Variation
(+ 3 Days)
+$80
Additional Scenario
+$25
Additional Graph/Chart
+$15
Model Documentation
+$50
Data Source Connectivity
(+ 3 Days)
+$120
Source Code
+$49Frequently asked questions
About Abdul
AI/ML Engineer | Computer Vision, Agentic AI & Lean Automation
Lahore, Pakistan - 11:48 pm local time
Computer Vision (The "Eyes"):
— Multi-stage pipelines combining YOLO (v8-v11), RetinaFace, DeepFace, and MediaPipe for detection, face verification, and liveness/anti-spoofing
— Few-shot classification via CLIP embeddings and clustering, built a system that classifies fine-grained categories using embedding similarity rather than large labeled datasets
— Prior work with ControlNet and OpenPose for pose-guided image manipulation
Agentic AI & Automation (The "Hands"):
— Voice AI agents (VAPI, ElevenLabs) handling real inbound/outbound calls, checking live calendar availability, and completing bookings end to end
— Workflow automation with n8n, connecting AI models to real business tools
— Backend architecture for LLM-powered systems: FastAPI, PostgreSQL, LangChain-based RAG pipelines, built with graceful fallback handling so provider failures never surface as raw errors to end users
Cost-Conscious ML:
— Weak supervision (Snorkel) for programmatic labeling instead of manual annotation
— Classical ML (XGBoost, Scikit-learn) where it beats an LLM call on tabular data, cheaper, faster, more interpretable
Stack: Python (daily), SQL, FastAPI, PostgreSQL, YOLO, CLIP, DeepFace, RetinaFace, MediaPipe, LangChain, n8n, VAPI, ElevenLabs, React.
I'm early in my professional career, but not new to shipping, I build production AI/backend systems daily, and everything above is work I've actually built and can walk you through in real detail, not a keyword list. If you need a CV pipeline that handles messy real-world input, or an agent that actually finishes tasks instead of just chatting, let's talk.
Steps for completing your project
After purchasing the project, send requirements so Abdul can start the project.
Delivery time starts when Abdul receives requirements from you.
Abdul works on your project following the steps below.
Revisions may occur after the delivery date.
Requirements & data review
review your data/use case, confirm accuracy targets and edge cases to handle.
Model selection & pipeline design
choose the right architecture (detection, verification, or multi-stage pipeline) for your accuracy/speed tradeoffs.