Blog
From the lab
Engineering deep-dives, product updates, and research notes from the team building visual intelligence.
What 1,000 Hours of Indian CCTV Footage Taught Us About Traffic
Vehicle distributions, temporal patterns, violation hotspots, and edge cases — the data tells a story that no traffic survey captures.
DrikNetra: Multi-Frame Super-Resolution for License Plates That Cameras Can't Read
Single frames from CCTV are too blurry. We fuse multiple video frames into one crisp plate image — turning 32x16 pixel smears into readable text.
drik-bench: If You Build the Benchmark, You Define What 'Good' Means
Why COCO and KITTI are useless for Indian traffic — and how we built a benchmark suite that measures what actually matters.
Edge-First: Our Real-Time Traffic Intelligence Architecture
How we process 200+ camera feeds at 30 FPS with sub-100ms latency — from RTSP ingestion to scene reasoning — without touching the cloud.
Building DrikSynth: Synthetic Data for Traffic Scenes That Don't Exist Yet
How we use Unreal Engine to generate 100K annotated traffic scenes — covering edge cases that would take years to collect in the real world.
Indian Traffic Is the Hardest Computer Vision Problem on Earth
50 million cameras. 300 million vehicles. 50+ vehicle types. No lane discipline. Why existing datasets fail — and why solving Indian traffic means solving everything.
Open-Sourcing DrikLabel: AI-Assisted Video Annotation
Why we're giving away our annotation tool — and how active learning propagation makes video labeling 10x faster.
Building a Visual Reasoning Engine for Indian Traffic
How we designed a five-level reasoning pipeline that goes from pixel-level detection to predictive intelligence — and why Indian roads are the ultimate proving ground.