# madhu-basavanna/computer_vision architecture

> Architecture diagram of the GitHub repository madhu-basavanna/computer_vision, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-02). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/madhu-basavanna/computer_vision
- Repository: https://github.com/madhu-basavanna/computer_vision

## Overview

The repository is an experimental vision pipeline for learning self-supervised image features and using them for semantic or class-conditioned segmentation, with a separate nearest-neighbor analysis workflow. Prepared image crops or annotated COCO data enter dataset readers and training scripts; the resulting models produce pixel masks and segmentation metrics. The README documents the principal pipeline and capabilities. Some subsystem links rely on that documentation because the relevant source was not excerpted; those limits are reflected in the evidence citations.

Main entry points: [`dt_multiclass_ss.py`](https://github.com/madhu-basavanna/computer_vision/blob/master/dt_multiclass_ss.py) (Multiclass training), [`dt_single_ss.py`](https://github.com/madhu-basavanna/computer_vision/blob/master/dt_single_ss.py) (Class-conditioned training), [`pretrain.py`](https://github.com/madhu-basavanna/computer_vision/blob/master/pretrain.py) (DINO training), [`models/pretraining_backbone.py`](https://github.com/madhu-basavanna/computer_vision/blob/master/models/pretraining_backbone.py) (ResNet feature backbone)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_data["Data preparation"]
  node_pretrain_data["Crop reader<br/>[pretraining.py]"]
  node_seg_data["Annotation readers<br/>[segmentation.py]"]
  node_augment["Crop augmentations"]
  node_transforms["Image transforms<br/>[transforms.py]"]
end

subgraph group_learning["Representation learning"]
  node_pretrain["DINO training<br/>[pretrain.py]"]
  node_backbone["ResNet feature backbone"]
end

subgraph group_segmentation["Segmentation"]
  node_multi_train["Multiclass training"]
  node_single_train["Class-conditioned training<br/>[dt_single_ss.py]"]
  node_segmentator["Multiclass segmentator"]
  node_attention_model["Attention segmentator"]
  node_attention["Class-conditioned attention<br/>[attention_layer.py]"]
  node_decoder["DeepLab decoder<br/>[deeplab_decoder.py]"]
end

subgraph group_analysis["Analysis and support"]
  node_retrieval["Nearest-neighbor analysis"]
  node_results["Masks and mIoU"]
  node_weights["Checkpoint loading<br/>[weights.py]"]
  node_meters["Training meters"]
  node_meter_base["Meter base<br/>[meter.py]"]
end

node_inputs["Prepared datasets"]

node_inputs -->|"provides crops"| node_pretrain_data
node_inputs -->|"provides annotations"| node_seg_data
node_pretrain -->|"loads images"| node_pretrain_data
node_pretrain -->|"applies crops"| node_augment
node_pretrain -->|"trains networks"| node_backbone
node_pretrain -->|"exports analysis"| node_results
node_retrieval -->|"extracts features"| node_backbone
node_retrieval -->|"reads crops"| node_pretrain_data
node_multi_train -->|"loads masks"| node_seg_data
node_multi_train -->|"prepares inputs"| node_transforms
node_multi_train -->|"uses features"| node_backbone
node_multi_train -->|"trains model"| node_segmentator
node_segmentator -->|"decodes features"| node_decoder
node_multi_train -->|"loads checkpoint"| node_weights
node_multi_train -->|"tracks training"| node_meters
node_multi_train -->|"evaluates masks"| node_results
node_single_train -->|"loads masks"| node_seg_data
node_single_train -->|"prepares inputs"| node_transforms
node_single_train -->|"uses features"| node_backbone
node_single_train -->|"trains model"| node_attention_model
node_attention_model -->|"applies attention"| node_attention
node_attention_model -->|"decodes features"| node_decoder
node_single_train -->|"loads checkpoint"| node_weights
node_single_train -->|"tracks training"| node_meters
node_single_train -->|"evaluates masks"| node_results
node_meters -->|"extends"| node_meter_base

click node_pretrain_data "https://github.com/madhu-basavanna/computer_vision/blob/master/data/pretraining.py"
click node_seg_data "https://github.com/madhu-basavanna/computer_vision/blob/master/data/segmentation.py"
click node_augment "https://github.com/madhu-basavanna/computer_vision/blob/master/data/data_augmentation.py"
click node_transforms "https://github.com/madhu-basavanna/computer_vision/blob/master/data/transforms.py"
click node_pretrain "https://github.com/madhu-basavanna/computer_vision/blob/master/pretrain.py"
click node_backbone "https://github.com/madhu-basavanna/computer_vision/blob/master/models/pretraining_backbone.py"
click node_retrieval "https://github.com/madhu-basavanna/computer_vision/blob/master/nearest_neighbors.py"
click node_multi_train "https://github.com/madhu-basavanna/computer_vision/blob/master/dt_multiclass_ss.py"
click node_single_train "https://github.com/madhu-basavanna/computer_vision/blob/master/dt_single_ss.py"
click node_segmentator "https://github.com/madhu-basavanna/computer_vision/blob/master/models/second_segmentation.py"
click node_attention_model "https://github.com/madhu-basavanna/computer_vision/blob/master/models/att_segmentation.py"
click node_attention "https://github.com/madhu-basavanna/computer_vision/blob/master/models/attention_layer.py"
click node_decoder "https://github.com/madhu-basavanna/computer_vision/blob/master/models/deeplab_decoder.py"
click node_weights "https://github.com/madhu-basavanna/computer_vision/blob/master/utils/weights.py"
click node_meters "https://github.com/madhu-basavanna/computer_vision/blob/master/utils/meters/averagevaluemeter.py"
click node_meter_base "https://github.com/madhu-basavanna/computer_vision/blob/master/utils/meters/meter.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_pretrain_data,node_seg_data,node_augment,node_transforms toneBlue
class node_pretrain,node_backbone toneAmber
class node_multi_train,node_single_train,node_segmentator,node_attention_model,node_attention,node_decoder toneMint
class node_retrieval,node_results,node_weights,node_meters,node_meter_base toneRose
class node_inputs toneIndigo
```

## Components

### Data preparation

- [Crop reader](https://github.com/madhu-basavanna/computer_vision/blob/master/data/pretraining.py): `data/pretraining.py`
- [Annotation readers](https://github.com/madhu-basavanna/computer_vision/blob/master/data/segmentation.py): `data/segmentation.py`
- [Crop augmentations](https://github.com/madhu-basavanna/computer_vision/blob/master/data/data_augmentation.py): `data/data_augmentation.py`
- [Image transforms](https://github.com/madhu-basavanna/computer_vision/blob/master/data/transforms.py): `data/transforms.py`

### Representation learning

- [DINO training](https://github.com/madhu-basavanna/computer_vision/blob/master/pretrain.py): `pretrain.py`
- [ResNet feature backbone](https://github.com/madhu-basavanna/computer_vision/blob/master/models/pretraining_backbone.py): `models/pretraining_backbone.py`

### Segmentation

- [Multiclass training](https://github.com/madhu-basavanna/computer_vision/blob/master/dt_multiclass_ss.py): `dt_multiclass_ss.py`
- [Class-conditioned training](https://github.com/madhu-basavanna/computer_vision/blob/master/dt_single_ss.py): `dt_single_ss.py`
- [Multiclass segmentator](https://github.com/madhu-basavanna/computer_vision/blob/master/models/second_segmentation.py): `models/second_segmentation.py`
- [Attention segmentator](https://github.com/madhu-basavanna/computer_vision/blob/master/models/att_segmentation.py): `models/att_segmentation.py`
- [Class-conditioned attention](https://github.com/madhu-basavanna/computer_vision/blob/master/models/attention_layer.py): `models/attention_layer.py`
- [DeepLab decoder](https://github.com/madhu-basavanna/computer_vision/blob/master/models/deeplab_decoder.py): `models/deeplab_decoder.py`

### Analysis and support

- [Nearest-neighbor analysis](https://github.com/madhu-basavanna/computer_vision/blob/master/nearest_neighbors.py): `nearest_neighbors.py`
- **Masks and mIoU**
- [Checkpoint loading](https://github.com/madhu-basavanna/computer_vision/blob/master/utils/weights.py): `utils/weights.py`
- [Training meters](https://github.com/madhu-basavanna/computer_vision/blob/master/utils/meters/averagevaluemeter.py): `utils/meters/averagevaluemeter.py`
- [Meter base](https://github.com/madhu-basavanna/computer_vision/blob/master/utils/meters/meter.py): `utils/meters/meter.py`

### Other components

- **Prepared datasets**

## Connections

- Prepared datasets → Crop reader: provides crops (evidence: `README.md`)
- Prepared datasets → Annotation readers: provides annotations (evidence: `README.md`)
- DINO training → Crop reader: loads images (evidence: `pretrain.py`)
- DINO training → Crop augmentations: applies crops (evidence: `pretrain.py`)
- DINO training → ResNet feature backbone: trains networks (evidence: `pretrain.py`)
- DINO training → Masks and mIoU: exports analysis (evidence: `README.md`)
- Nearest-neighbor analysis → ResNet feature backbone: extracts features (evidence: `README.md`)
- Nearest-neighbor analysis → Crop reader: reads crops (evidence: `README.md`)
- Multiclass training → Annotation readers: loads masks (evidence: `dt_multiclass_ss.py`)
- Multiclass training → Image transforms: prepares inputs (evidence: `dt_multiclass_ss.py`)
- Multiclass training → ResNet feature backbone: uses features (evidence: `dt_multiclass_ss.py`)
- Multiclass training → Multiclass segmentator: trains model (evidence: `dt_multiclass_ss.py`)
- Multiclass segmentator → DeepLab decoder: decodes features (evidence: `models/second_segmentation.py`)
- Multiclass training → Checkpoint loading: loads checkpoint (evidence: `dt_multiclass_ss.py`)
- Multiclass training → Training meters: tracks training (evidence: `dt_multiclass_ss.py`)
- Multiclass training → Masks and mIoU: evaluates masks (evidence: `README.md`)
- Class-conditioned training → Annotation readers: loads masks (evidence: `dt_single_ss.py`)
- Class-conditioned training → Image transforms: prepares inputs (evidence: `dt_single_ss.py`)
- Class-conditioned training → ResNet feature backbone: uses features (evidence: `dt_single_ss.py`)
- Class-conditioned training → Attention segmentator: trains model (evidence: `dt_single_ss.py`)
- Attention segmentator → Class-conditioned attention: applies attention (evidence: `models/att_segmentation.py`)
- Attention segmentator → DeepLab decoder: decodes features (evidence: `models/att_segmentation.py`)
- Class-conditioned training → Checkpoint loading: loads checkpoint (evidence: `dt_single_ss.py`)
- Class-conditioned training → Training meters: tracks training (evidence: `dt_single_ss.py`)
- Class-conditioned training → Masks and mIoU: evaluates masks (evidence: `README.md`)
- Training meters → Meter base: extends (evidence: `utils/meters/averagevaluemeter.py`)
