# pranavv0307/rs-uep-jscc architecture

> Architecture diagram of the GitHub repository pranavv0307/rs-uep-jscc, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-05). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/pranavv0307/rs-uep-jscc
- Repository: https://github.com/pranavv0307/rs-uep-jscc

## Overview

This research codebase studies image transmission over noisy or erased channels, with ADJSCC-Q plus Reed–Solomon unequal error protection as its principal path: load CIFAR-10, encode and quantize images, rank latent symbols by importance, protect and transmit packets, reconstruct images, and evaluate quality. Training, clean evaluation, AWGN demos, and erasure studies are separate workflows around the model and coding stages. Edges follow the shown excerpts, SOURCE INDEX references, or documented behavior; unsampled subsystem details are not inferred.

Main entry points: [`experiments/toy_demo.py`](https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/toy_demo.py) (AWGN demo), [`experiments/eval_rs_uep.py`](https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/eval_rs_uep.py) (RS-UEP evaluation), [`experiments/train_adjsccq_train_split.py`](https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/train_adjsccq_train_split.py) (ADJSCC-Q training), [`coding/rs_pipeline.py`](https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/rs_pipeline.py) (RS tier pipeline)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_experiments["Training &amp; Evaluation"]
  node_train["ADJSCC-Q training"]
  node_clean_eval["Clean evaluation"]
  node_rs_eval["RS-UEP evaluation<br/>[eval_rs_uep.py]"]
  node_metrics["Image metrics<br/>[metrics.py]"]
  node_toy["AWGN demo<br/>[toy_demo.py]"]
end

subgraph group_models["Image Codecs"]
  node_adjsccq["Quantized ADJSCC<br/>[adjsccq.py]"]
  node_adjscc["ADJSCC backbone<br/>[adjscc.py]"]
  node_deepjscc["DeepJSCC baseline<br/>[deepjscc.py]"]
  node_attention["Attention module<br/>[Attention.py]"]
  node_quantizer["Scalar quantizer<br/>[quantizer.py]"]
  node_awgn["AWGN channel<br/>[channel.py]"]
end

subgraph group_coding["Packet Protection"]
  node_importance["Importance ranking<br/>[importance.py]"]
  node_pipeline["RS tier pipeline<br/>[rs_pipeline.py]"]
  node_packetizer["Packetizer<br/>[packetizer.py]"]
  node_erasure["Erasure channel<br/>[erasure_channel.py]"]
  node_reed_solomon["Reed–Solomon<br/>[reed_solomon.py]"]
  node_tiering["Tier allocation<br/>[tiering.py]"]
end

subgraph group_data["Image Data"]
  node_cifar["CIFAR-10 loader<br/>[cifar10.py]"]
end

node_user(("Researcher"))

node_user -->|"starts training"| node_train
node_user -->|"evaluates codec"| node_clean_eval
node_user -->|"tests erasures"| node_rs_eval
node_cifar -->|"provides images"| node_train
node_cifar -->|"provides images"| node_clean_eval
node_train -->|"trains model"| node_adjsccq
node_train -->|"computes quality"| node_metrics
node_train -->|"evaluates protection"| node_pipeline
node_clean_eval -->|"runs codec"| node_adjsccq
node_clean_eval -->|"scores reconstructions"| node_metrics
node_rs_eval -->|"loads images"| node_cifar
node_rs_eval -->|"evaluates packet coding"| node_pipeline
node_rs_eval -->|"scores reconstructions"| node_metrics
node_adjsccq -->|"wraps backbone"| node_adjscc
node_adjsccq -->|"quantizes latents"| node_quantizer
node_adjscc -->|"uses attention"| node_attention
node_toy -->|"loads images"| node_cifar
node_toy -->|"runs encoder"| node_adjscc
node_toy -->|"runs baseline"| node_deepjscc
node_toy -->|"transmits latents"| node_awgn
node_toy -->|"scores reconstructions"| node_metrics
node_pipeline -->|"orders symbols"| node_importance
node_pipeline -->|"forms packets"| node_packetizer
node_pipeline -->|"encodes tiers"| node_reed_solomon
node_erasure -->|"erases packets"| node_packetizer
node_rs_eval -->|"simulates packet loss"| node_erasure

click node_cifar "https://github.com/pranavv0307/rs-uep-jscc/blob/main/data/cifar10.py"
click node_train "https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/train_adjsccq_train_split.py"
click node_clean_eval "https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/eval_adjsccq_clean.py"
click node_rs_eval "https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/eval_rs_uep.py"
click node_metrics "https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/metrics.py"
click node_toy "https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/toy_demo.py"
click node_adjsccq "https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/adjsccq.py"
click node_adjscc "https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/adjscc.py"
click node_deepjscc "https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/deepjscc.py"
click node_attention "https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/Attention.py"
click node_quantizer "https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/quantizer.py"
click node_awgn "https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/channel.py"
click node_importance "https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/importance.py"
click node_pipeline "https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/rs_pipeline.py"
click node_packetizer "https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/packetizer.py"
click node_erasure "https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/erasure_channel.py"
click node_reed_solomon "https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/reed_solomon.py"
click node_tiering "https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/tiering.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_train,node_clean_eval,node_rs_eval,node_metrics,node_toy toneBlue
class node_adjsccq,node_adjscc,node_deepjscc,node_attention,node_quantizer,node_awgn toneAmber
class node_importance,node_pipeline,node_packetizer,node_erasure,node_reed_solomon,node_tiering toneMint
class node_cifar toneRose
class node_user toneIndigo
```

## Components

### Training & Evaluation

- [ADJSCC-Q training](https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/train_adjsccq_train_split.py): `experiments/train_adjsccq_train_split.py`
- [Clean evaluation](https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/eval_adjsccq_clean.py): `experiments/eval_adjsccq_clean.py`
- [RS-UEP evaluation](https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/eval_rs_uep.py): `experiments/eval_rs_uep.py`
- [Image metrics](https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/metrics.py): `experiments/metrics.py`
- [AWGN demo](https://github.com/pranavv0307/rs-uep-jscc/blob/main/experiments/toy_demo.py): `experiments/toy_demo.py`

### Image Codecs

- [Quantized ADJSCC](https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/adjsccq.py): `models/adjsccq.py`
- [ADJSCC backbone](https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/adjscc.py): `models/adjscc.py`
- [DeepJSCC baseline](https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/deepjscc.py): `models/deepjscc.py`
- [Attention module](https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/Attention.py): `models/Attention.py`
- [Scalar quantizer](https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/quantizer.py): `models/quantizer.py`
- [AWGN channel](https://github.com/pranavv0307/rs-uep-jscc/blob/main/models/channel.py): `models/channel.py`

### Packet Protection

- [Importance ranking](https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/importance.py): `coding/importance.py`
- [RS tier pipeline](https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/rs_pipeline.py): `coding/rs_pipeline.py`
- [Packetizer](https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/packetizer.py): `coding/packetizer.py`
- [Erasure channel](https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/erasure_channel.py): `coding/erasure_channel.py`
- [Reed–Solomon](https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/reed_solomon.py): `coding/reed_solomon.py`
- [Tier allocation](https://github.com/pranavv0307/rs-uep-jscc/blob/main/coding/tiering.py): `coding/tiering.py`

### Image Data

- [CIFAR-10 loader](https://github.com/pranavv0307/rs-uep-jscc/blob/main/data/cifar10.py): `data/cifar10.py`

### Other components

- **Researcher**

## Connections

- Researcher → ADJSCC-Q training: starts training
- Researcher → Clean evaluation: evaluates codec
- Researcher → RS-UEP evaluation: tests erasures
- CIFAR-10 loader → ADJSCC-Q training: provides images (evidence: `experiments/train_adjsccq_train_split.py`)
- CIFAR-10 loader → Clean evaluation: provides images (evidence: `experiments/eval_adjsccq_clean.py`)
- ADJSCC-Q training → Quantized ADJSCC: trains model (evidence: `experiments/train_adjsccq_train_split.py`)
- ADJSCC-Q training → Image metrics: computes quality (evidence: `experiments/train_adjsccq_train_split.py`)
- ADJSCC-Q training → RS tier pipeline: evaluates protection (evidence: `experiments/train_adjsccq_train_split.py`)
- Clean evaluation → Quantized ADJSCC: runs codec (evidence: `experiments/eval_adjsccq_clean.py`)
- Clean evaluation → Image metrics: scores reconstructions (evidence: `experiments/eval_adjsccq_clean.py`)
- RS-UEP evaluation → CIFAR-10 loader: loads images (evidence: `experiments/eval_rs_uep.py`)
- RS-UEP evaluation → RS tier pipeline: evaluates packet coding (evidence: `experiments/eval_rs_uep.py`)
- RS-UEP evaluation → Image metrics: scores reconstructions (evidence: `experiments/eval_rs_uep.py`)
- Quantized ADJSCC → ADJSCC backbone: wraps backbone (evidence: `models/adjsccq.py`)
- Quantized ADJSCC → Scalar quantizer: quantizes latents (evidence: `models/adjsccq.py`)
- ADJSCC backbone → Attention module: uses attention (evidence: `models/adjscc.py`)
- AWGN demo → CIFAR-10 loader: loads images (evidence: `experiments/toy_demo.py`)
- AWGN demo → ADJSCC backbone: runs encoder (evidence: `experiments/toy_demo.py`)
- AWGN demo → DeepJSCC baseline: runs baseline (evidence: `experiments/toy_demo.py`)
- AWGN demo → AWGN channel: transmits latents (evidence: `experiments/toy_demo.py`)
- AWGN demo → Image metrics: scores reconstructions (evidence: `experiments/toy_demo.py`)
- RS tier pipeline → Importance ranking: orders symbols (evidence: `coding/rs_pipeline.py`)
- RS tier pipeline → Packetizer: forms packets (evidence: `coding/rs_pipeline.py`)
- RS tier pipeline → Reed–Solomon: encodes tiers (evidence: `coding/rs_pipeline.py`)
- Erasure channel → Packetizer: erases packets (evidence: `coding/erasure_channel.py`)
- RS-UEP evaluation → Erasure channel: simulates packet loss (evidence: `experiments/eval_rs_uep.py`)
