Rico贾若童的博客

[BEV] Feature Pyramid Network

Posted by Rico's Nerd Cluster on April 10, 2026

[BEV] BEV Ideas and View Transformation

Training pipelines, multiview geometry, and how to lift image features into BEV

1. The Core Goal of a BEV Network A Bird’s Eye View (BEV) network aims to reconstruct, in real time using only onboard cameras, a local 3D map of comparable quality to what an offline reconstructi...

Posted by Rico's Nerd Cluster on April 8, 2026

[BEV] BEV Introduction: Tesla's Challenges and Architecture

Why per-camera detection falls short, and how BEV solves it

1. Tesla’s Perception Challenges (2021) Before BEV, Tesla’s pipeline detected objects and lanes independently in each camera view and then tried to fuse the results. This created fundamental probl...

Posted by Rico's Nerd Cluster on April 5, 2026

[BEV] BEV Introduction

Understanding L2 and L2+ Perception: Why the Real Challenge Is System Design, Not Just Better Detection When people talk about assisted driving, they often jump straight to compute power, percepti...

Posted by Rico's Nerd Cluster on April 2, 2026

MPPI And Motion Control On My Diff Drive

Remote Claude Code Setup

Motor Control So, tested on carpet, I think it’s about not having (+/-) on the spot that makes it motor controller - aware that the robot turns better using something like: left = ±u righ...

Posted by Rico's Nerd Cluster on March 25, 2026

AI Agent Setup

Remote Claude Code Setup

Remote Claude Code Setup Bring up tailscale: sudo tailscale up Bring up code-server: code-server --bind-addr $(tailscale ip -4):8081 To see password for access cat ~/.config/code-server/con...

Posted by Rico's Nerd Cluster on March 22, 2026

Model Parameter vs VRAM

VRAM

VRAM = Video RAM. It is the memory on your GPU, separate from your normal computer RAM. The GPU uses VRAM to store: model weights input images / batches intermediate activations gradients...

Posted by Rico's Nerd Cluster on March 5, 2026

[Point Cloud Compression] Draco

Introduction Draco is Google’s open-source library for compressing 3D geometric meshes and point clouds. It quantizes floating-point attributes (positions, normals, colors, texture coordinates) in...

Posted by Rico's Nerd Cluster on March 2, 2026

[ML] HuggingFace Trainer

Introduction The Trainer class provides a feature-complete training loop for PyTorch, supporting distributed training on multiple GPUs/TPUs and mixed precision via NVIDIA Apex, AMD ROCm, and torch...

Posted by Rico's Nerd Cluster on February 28, 2026

[ML] Model Visualization

ONNX, Netron, Model Size

Netron Netron is a viewer for neural network models. The typical workflow is to export a PyTorch model to ONNX format and then drag the .onnx file into the Netron web app. What is ONNX? Open Neur...

Posted by Rico's Nerd Cluster on February 26, 2026

Rico's Nerd Cluster

[BEV] Feature Pyramid Network

[BEV] BEV Ideas and View Transformation

Training pipelines, multiview geometry, and how to lift image features into BEV

[BEV] BEV Introduction: Tesla's Challenges and Architecture

Why per-camera detection falls short, and how BEV solves it

[BEV] BEV Introduction

MPPI And Motion Control On My Diff Drive

Remote Claude Code Setup

AI Agent Setup

Remote Claude Code Setup

Model Parameter vs VRAM

VRAM

[Point Cloud Compression] Draco

[ML] HuggingFace Trainer

[ML] Model Visualization

ONNX, Netron, Model Size

FEATURED TAGS

ABOUT ME

FRIENDS