Satya Mallick

Satya Mallick

Creator
0 followers

CEO, https://t.co/CzUdJlxzJM. Course Director, https://t.co/O2Tz9vUOQ8 Entrepreneur. Ph.D. ( Computer Vision & Machine Learning ). Author: https://t.co/olraDEG5Ue

Write Vision Code Once, Run Anywhere with OpenCV5 HAL
SocialJun 16, 2026

Write Vision Code Once, Run Anywhere with OpenCV5 HAL

Write your computer vision code once. Run it fast on a CPU, a GPU, or a dedicated AI accelerator — without changing a single line. That's the promise of the Hardware Abstraction Layer (HAL) in OpenCV 5. Part 1 https://t.co/hzs5CNr1kd #OpenCV5 #ComputerVision...

By Satya Mallick
Pointing Ability Boosts AI Agent Speed and Accuracy
SocialJun 16, 2026

Pointing Ability Boosts AI Agent Speed and Accuracy

Part 2: The future of AI agents may depend on something basic — not just whether they can see, but whether they can point. LocateAnything-3B: 12.7 boxes/sec, ~2.5x faster than Rex-Omni, and more accurate. Speed + precision is the whole game. https://t.co/HZQ19AwHYk...

By Satya Mallick
Speed Makes NVIDIA's LocateAnything Truly Actionable AI
SocialJun 12, 2026

Speed Makes NVIDIA's LocateAnything Truly Actionable AI

Part 1: NVIDIA's LocateAnything is built for the moment AI stops answering questions and starts pointing, clicking, reading, and acting. Speed isn't a luxury — it's the difference between a useful agent and a confused one. https://t.co/kHFOKMDYOB https://t.co/lyEEc5ADUB

By Satya Mallick
OpenCV 5 DNN Engine Boosts ONNX Support, Speed
SocialJun 12, 2026

OpenCV 5 DNN Engine Boosts ONNX Support, Speed

The Three DNN Engines of OpenCV 5. The old 4.x DNN engine imported ~22% of ONNX. The new graph-based engine pushes past 80%, fuses MatMul→Softmax→MatMul into one FlashAttention layer, and runs YOLO26n 41% faster than ONNX Runtime — no code changes. Deep...

By Satya Mallick
YOLOE-26 Enables Open‑Vocabulary Detection via Text, Visual, or No Prompt
SocialJun 1, 2026

YOLOE-26 Enables Open‑Vocabulary Detection via Text, Visual, or No Prompt

YOLOE-26 turns object detection into three ways of saying "find this": → Text prompt (name it) → Visual prompt (show it) → Prompt-free (let the model decide) Closed-set rigidity → open-vocabulary conversation. Tutorial + benchmarks: https://t.co/od9zkfvMaX https://t.co/t1bXsIl1HJ

By Satya Mallick
YOLOE Enables Zero‑Overhead Open‑Vocabulary Detection
SocialMay 29, 2026

YOLOE Enables Zero‑Overhead Open‑Vocabulary Detection

Object detection is shifting from "models that recognize fixed categories" to "models that understand concepts described in language." YOLOE delivers open-vocabulary detection at full YOLO speed — text module fused into the head, zero runtime overhead. Full tutorial + code: https://t.co/JKzMcctoGw

By Satya Mallick
Robotic “Eyeball” Ensures Vision Pro Quality
SocialMay 26, 2026

Robotic “Eyeball” Ensures Vision Pro Quality

This robot's only job is to pretend it's your eyeball 👁️🤖 At Display Week 2026, Dr. Satya Mallick visits Gamma Scientific — the 6-axis robot AR/VR brands use to QA every headset before launch. 18+ tests in one rig: contrast, parallax,...

By Satya Mallick
MoE Training Amplifies Tiny Expertise Into Mastery
SocialMay 22, 2026

MoE Training Amplifies Tiny Expertise Into Mastery

MoE Training, Part 2 — in one tweet: You start with random weights. By chance, one expert is slightly better at legal questions. Router notices, sends more its way. It gets better. Snowballs. Same compounding loop that turns a slightly-talented 7-year-old into...

By Satya Mallick
Specialization Emerges Naturally in MoE Training
SocialMay 21, 2026

Specialization Emerges Naturally in MoE Training

MoE Training, Part 1 — in one tweet: You do NOT assign "this expert handles medicine, this one handles law." You start with 9 random experts + a router. The router learns to pick 2–3 per question. Specialization emerges from data, not...

By Satya Mallick
Frontier LLMs Adopt Mixture‑of‑Experts for Efficient Compute
SocialMay 15, 2026

Frontier LLMs Adopt Mixture‑of‑Experts for Efficient Compute

Why every frontier LLM is converging on Mixture of Experts 🧵 Trillion-parameter model. Single query. You don't need the whole thing. A router picks a subset of "experts." Medical question → medical expert. Legal → legal. Some models keep one generalist always...

By Satya Mallick
VLMs Share Label Role, Differ Vastly in Capabilities
SocialMay 15, 2026

VLMs Share Label Role, Differ Vastly in Capabilities

"VLM" is doing a lot of heavy lifting as a label. CLIP → image-text alignment, zero-shot recognition Moondream → grounding ("find the guy in red") Qwen3-VL → agentic + GUI + long video understanding Same category. Wildly different tools. Dr. Satya Mallick explains → https://t.co/4AZvwlbKDm #VLM...

By Satya Mallick
YOLO26‑Seg Delivers Razor‑sharp Masks, 43% CPU Speed Boost
SocialMay 14, 2026

YOLO26‑Seg Delivers Razor‑sharp Masks, 43% CPU Speed Boost

Pt. 2 — YOLO26-Seg is wild: → Distribution Focal Loss removed → MuSGD optimizer (hybrid borrowed from LLM training) → NMS baked into the model → Boundary-aware supervision = razor-sharp masks → Up to 43% faster on CPU → One ONNX export → Pi, drone, phone Deep...

By Satya Mallick
Real‑Time Monocular Depth From One Camera: Depth Anything V2
SocialMay 9, 2026

Real‑Time Monocular Depth From One Camera: Depth Anything V2

What if accurate depth maps could be generated from a single RGB image — without LiDAR or stereo cameras? That’s exactly what Depth Anything V2 achieves. In 2024, monocular depth estimation reached a major breakthrough: ✔ Fast ✔ Lightweight ✔ Temporally stable ✔ Edge-device friendly Instead of...

By Satya Mallick
Experts Use Regularization; Novices Skip Its Top Benefits
SocialMay 7, 2026

Experts Use Regularization; Novices Skip Its Top Benefits

The four benefits in order of impact: 1. Prevents overfitting (the big one) 2. Adversarial robustness 3. Augments small datasets 4. Softer decision boundaries Used by experts. Skipped by most novices. Don't be a novice. https://t.co/eK6lhglg6o

By Satya Mallick
Satya Mallick | Pulse