# pengyizhang/slimyolov3 architecture

> Architecture diagram of the GitHub repository pengyizhang/slimyolov3, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2025-06-16). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/pengyizhang/slimyolov3
- Repository: https://github.com/pengyizhang/slimyolov3

## Overview

##### Project Type and Purpose
- Type: Machine-learning training & inference pipeline, combining Darknet (C) and PyTorch (Python) components.
- Purpose: Learn a sparse (“slim”) YOLOv3 object detector via channel pruning, deployable on resource-constrained UAVs.

##### Main Architectural Components
A. Data Management
– VisDrone2019 Dataset (images, labels, train/val lists) in VisDrone2019/
– Data‐loader configurations in VisDrone2019/drone.data and cfg/*.cfg
B. Configuration Layer
– Unpruned & pruned network definitions in cfg/00-unpruned, cfg/1iter-pruned, …
– YOLOv3 model configs (yolov3.cfg, yolov3-spp3.cfg, yolov3-tiny.cfg)
C. Training Engines
1. Darknet C Binary
– Normal training: `./darknet detector train ...`
– Fine‐tuning pruned networks: `./darknet detector train cfg/prune_*.cfg ...`
2. PyTorch Scripts (ultralytics/yolov3)
– Sparsity training: `yolov3/train_drone.py` with L1‐regularization
– Pruning: `yolov3/prune.py` and root `prune.py` (collect channel‐scaling factors, prune)
D. Pruning Workflow
– Input: sparsity-trained weights + config
– Process: channel selection & config rewriting → pruned CFGs + pruned weight files
E. Evaluation & Results
– Darknet inference: `darknet detector valid …`
– PyTorch inference: `yolov3/test_drone.py`, `test_single_image.py`
– Outputs: accuracy metrics, demo video (results/demo.mp4), images directory

##### Relationships & Data Flow
- Data feeds into both Darknet and PyTorch training engines.
- Each engine reads .cfg and .data files, emits weight files.
- Sparsity training outputs “sparsity.weights” → fed into pruning scripts.
- Pruning scripts read granular cfgs, produce pruned cfgs in cfg/** and pruned‐weights.
- Fine-tuning uses pruned cfgs + pruned weights → final pruned weights in backup directories.
- Evaluation modules consume final weights + configs → produce metrics, demo.mp4.

##### Architectural Patterns & Design Principles
- Pipeline pattern: a linear sequence (Train → Sparsity → Prune → Fine-tune → Evaluate)
- Config-driven design: behavior controlled by .cfg files, parameters for sparsity, prune ratios
- Modular separation: C‐based Darknet vs Python/PyTorch modules, clear interfacing via configs/weights
- Extensibility: new pruning ratios via directory cfg/***, new datasets via VisDrone2019/*

##### Technologies & Frameworks
- Darknet (C, GPU accelerated) for core YOLO training/inference
- PyTorch (Python ≥1.0), ultralytics/yolov3 repo for sparsity training & Python-based pruning
- Python scripts: prune.py, sparsity.py (root and yolov3/)
- Data/config files: .data, .cfg, label lists

## Diagram (Mermaid)

```mermaid
flowchart TB
    %% Data Management
    subgraph "Data Management"
        direction TB
        Dataset[(VisDrone2019 Dataset)]:::data
        drone_data["drone.data (.data file)"]:::config
        labels["labels.txt"]:::config
        train_list["train.lists"]:::config
        val_list["val.lists"]:::config
    end

    %% Configuration Layer
    subgraph "Configuration Layer"
        direction TB
        subgraph "Unpruned Configs"
            direction TB
            up_cfg_dir["cfg/00-unpruned/"]:::config
            up_cfg1["yolov3.cfg"]:::config
            up_cfg2["yolov3-spp1.cfg"]:::config
            up_cfg3["yolov3-spp3.cfg"]:::config
            up_cfg4["yolov3-tiny.cfg"]:::config
        end
        subgraph "1st Iteration Pruned Configs"
            direction TB
            p1_cfg_dir["cfg/1iter-pruned/"]:::config
            p1_cfg1["prune_0.5.cfg"]:::config
            p1_cfg2["prune_0.9.cfg"]:::config
            p1_cfg3["prune_0.95.cfg"]:::config
        end
        subgraph "2nd Iteration Pruned Configs"
            direction TB
            p2_cfg_dir["cfg/2iter-pruned/"]:::config
            p2_cfg1["prune_0.5_0.5.cfg"]:::config
        end
        subgraph "3rd Iteration Pruned Configs"
            direction TB
            p3_cfg_dir["cfg/3iter-pruned/"]:::config
            p3_cfg1["prune_0.5_0.5_0.7.cfg"]:::config
        end
    end

    %% Training Engines
    subgraph "Training Engines"
        direction TB
        Darknet_Train["Darknet Trainer"]:::darknet
        PyTorch_Sparsity["PyTorch Sparsity Trainer"]:::pyTorch
    end

    %% Pruning Engine
    Prune_Engine["Prune Engine"]:::pyTorch

    %% Fine-tune Trainer
    FineTune["Fine-tune Trainer (Darknet)"]:::darknet

    %% Evaluation
    subgraph "Evaluation & Results"
        direction TB
        Darknet_Eval["Darknet Evaluator"]:::darknet
        PyTorch_Eval["PyTorch Evaluator"]:::pyTorch
        Demo["demo.mp4"]:::evaluation
    end

    %% Weights artifacts
    Sparsity_Weights[(sparsity.weights)]:::weight
    Pruned_Weights[(pruned.weights)]:::weight
    Final_Weights[(final_pruned.weights)]:::weight

    %% Flows
    Dataset -->|provides data| Darknet_Train
    Dataset -->|provides data| PyTorch_Sparsity
    drone_data --> Darknet_Train
    drone_data --> PyTorch_Sparsity
    labels --> PyTorch_Sparsity
    train_list --> PyTorch_Sparsity
    val_list --> PyTorch_Sparsity

    up_cfg1 --> Darknet_Train
    up_cfg2 --> Darknet_Train
    up_cfg3 --> Darknet_Train
    up_cfg4 --> Darknet_Train

    up_cfg_dir --> PyTorch_Sparsity
    PyTorch_Sparsity -->|L1 reg| Sparsity_Weights
    Sparsity_Weights --> Prune_Engine
    prune.py --> Prune_Engine
    p1_cfg_dir --> Prune_Engine
    Prune_Engine --> Pruned_Weights
    Pruned_Weights --> FineTune
    p1_cfg1 --> FineTune
    p2_cfg1 --> FineTune
    p3_cfg1 --> FineTune

    FineTune --> Final_Weights
    Final_Weights --> Darknet_Eval
    Final_Weights --> PyTorch_Eval
    Darknet_Eval --> Demo
    PyTorch_Eval --> Demo

    %% Click events
    click Dataset "https://github.com/pengyizhang/slimyolov3/tree/master/VisDrone2019/"
    click drone_data "https://github.com/pengyizhang/slimyolov3/blob/master/VisDrone2019/drone.data"
    click labels "https://github.com/pengyizhang/slimyolov3/blob/master/VisDrone2019/drone/labels.txt"
    click train_list "https://github.com/pengyizhang/slimyolov3/blob/master/VisDrone2019/drone/train.lists"
    click val_list "https://github.com/pengyizhang/slimyolov3/blob/master/VisDrone2019/drone/val.lists"
    click up_cfg_dir "https://github.com/pengyizhang/slimyolov3/tree/master/cfg/00-unpruned/"
    click up_cfg1 "https://github.com/pengyizhang/slimyolov3/blob/master/cfg/00-unpruned/yolov3.cfg"
    click up_cfg2 "https://github.com/pengyizhang/slimyolov3/blob/master/cfg/00-unpruned/yolov3-spp1.cfg"
    click up_cfg3 "https://github.com/pengyizhang/slimyolov3/blob/master/cfg/00-unpruned/yolov3-spp3.cfg"
    click up_cfg4 "https://github.com/pengyizhang/slimyolov3/blob/master/cfg/00-unpruned/yolov3-tiny.cfg"
    click p1_cfg_dir "https://github.com/pengyizhang/slimyolov3/tree/master/cfg/1iter-pruned/"
    click p1_cfg1 "https://github.com/pengyizhang/slimyolov3/blob/master/cfg/1iter-pruned/prune_0.5.cfg"
    click p1_cfg2 "https://github.com/pengyizhang/slimyolov3/blob/master/cfg/1iter-pruned/prune_0.9.cfg"
    click p1_cfg3 "https://github.com/pengyizhang/slimyolov3/blob/master/cfg/1iter-pruned/prune_0.95.cfg"
    click p2_cfg_dir "https://github.com/pengyizhang/slimyolov3/tree/master/cfg/2iter-pruned/"
    click p2_cfg1 "https://github.com/pengyizhang/slimyolov3/blob/master/cfg/2iter-pruned/prune_0.5_0.5.cfg"
    click p3_cfg_dir "https://github.com/pengyizhang/slimyolov3/tree/master/cfg/3iter-pruned/"
    click p3_cfg1 "https://github.com/pengyizhang/slimyolov3/blob/master/cfg/3iter-pruned/prune_0.5_0.5_0.7.cfg"
    click Prune_Engine "https://github.com/pengyizhang/slimyolov3/blob/master/prune.py"
    click PyTorch_Sparsity "https://github.com/pengyizhang/slimyolov3/blob/master/sparsity.py"

    %% Styles
    classDef data fill:#ffe4b5,stroke:#cc9900,stroke-width:1px
    classDef config fill:#e0e0e0,stroke:#666,stroke-width:1px
    classDef darknet fill:#add8e6,stroke:#0000ff,stroke-width:1px
    classDef pyTorch fill:#98fb98,stroke:#228b22,stroke-width:1px
    classDef weight fill:#ffdead,stroke:#daa520,stroke-width:1px
    classDef evaluation fill:#f0e68c,stroke:#b8860b,stroke-width:1px
```
