---
license: mit
library_name: onnx
pipeline_tag: image-classification
tags:
- image-classification
- mnist
- convolutional-neural-network
- onnx
- onnxruntime
- pytorch
- dlab
datasets:
- mnist
metrics:
- accuracy
---

# MNIST CNN Classifier

This repository contains a validation-selected MNIST CNN digit classifier trained with [dlab](https://github.com/tsilva/dlab).

## Architecture

![MNIST CNN architecture](assets/architecture.png)

## Results

5-seed held-out test evaluation:

| metric | value |
|---|---:|
| test accuracy | 99.6140% ± 0.0802 pp |
| test loss | 0.14677 ± 0.00272 |
| best validation loss | 0.14282 ± 0.00138 |

Per-seed held-out test results:

| seed | W&B run | test accuracy | test loss | best validation loss |
|---:|---|---:|---:|---:|
| 1 | [`5um57rnu`](https://wandb.ai/tsilva/dlab/runs/5um57rnu) | 99.6100% | 0.14772 | 0.14124 |
| 2 | [`23f1frqb`](https://wandb.ai/tsilva/dlab/runs/23f1frqb) | 99.6800% | 0.14437 | 0.14317 |
| 3 | [`25yaaj1o`](https://wandb.ai/tsilva/dlab/runs/25yaaj1o) | 99.4800% | 0.15107 | 0.14403 |
| 4 | [`3rrlxghp`](https://wandb.ai/tsilva/dlab/runs/3rrlxghp) | 99.6300% | 0.14573 | 0.14150 |
| 5 | [`y51200ov`](https://wandb.ai/tsilva/dlab/runs/y51200ov) | 99.6700% | 0.14494 | 0.14418 |

The ONNX model was exported from the seed-1 checkpoint, which had the best validation loss in the final 5-seed evaluation sweep. Test metrics were not used for checkpoint selection and were logged in W&B sweep [`ikfs5ox8`](https://wandb.ai/tsilva/dlab/sweeps/ikfs5ox8).

## Model Details

- Dataset: MNIST
- Architecture: CNN
- Channels: `[32, 64, 128]`
- Convolutions per stage: `2`
- Batch normalization: enabled
- Dropout: `0.1`
- Optimizer: Adam
- Learning rate: `0.001`
- Weight decay: `0.0001`
- Scheduler: OneCycleLR
- Label smoothing: `0.02`
- Weight averaging: EMA
- Batch size: `512`
- Training augmentation: random affine rotation/translation/scale
- Early stopping: validation loss, patience `8`, min delta `0.0005`
- Source W&B run: [`5um57rnu`](https://wandb.ai/tsilva/dlab/runs/5um57rnu)
- Source W&B sweep: [`ikfs5ox8`](https://wandb.ai/tsilva/dlab/sweeps/ikfs5ox8)

## Input / Output

Use `model.onnx` for code-independent inference.

- Input name: `images`
- Input shape: `[batch, 1, 28, 28]`
- Input dtype: `float32`
- Output name: `logits`
- Output shape: `[batch, 10]`

Preprocessing:

- Convert image to grayscale.
- Resize to `28 x 28`.
- Scale pixel values to `[0, 1]`.
- Normalize with mean `0.1307` and standard deviation `0.3081`.
- Arrange the tensor as channels-first `[batch, 1, 28, 28]`.

## Usage

Install the runtime dependencies:

```bash
pip install huggingface_hub onnxruntime pillow numpy
```

Run inference with the ONNX model:

```python
import numpy as np
import onnxruntime as ort
from huggingface_hub import hf_hub_download
from PIL import Image

LABELS = {
    0: "0",
    1: "1",
    2: "2",
    3: "3",
    4: "4",
    5: "5",
    6: "6",
    7: "7",
    8: "8",
    9: "9",
}

model_path = hf_hub_download(
    repo_id="tsilva/mnist-classifier-cnn",
    filename="model.onnx",
)

image = Image.open("example.png").convert("L").resize((28, 28))
x = np.asarray(image, dtype=np.float32) / 255.0
x = (x - 0.1307) / 0.3081
x = x[None, None, :, :].astype(np.float32)

session = ort.InferenceSession(model_path, providers=["CPUExecutionProvider"])
logits = session.run(["logits"], {"images": x})[0]
prediction = int(logits.argmax(axis=1)[0])

print(prediction, LABELS[prediction])
```

## Labels

MNIST labels:

| id | label |
|---:|---|
| 0 | 0 |
| 1 | 1 |
| 2 | 2 |
| 3 | 3 |
| 4 | 4 |
| 5 | 5 |
| 6 | 6 |
| 7 | 7 |
| 8 | 8 |
| 9 | 9 |

## Files

- `model.onnx`: ONNX export of the validation-selected checkpoint. Prefer this file for portable inference.
- `model.ckpt`: PyTorch Lightning checkpoint for the same model. This is code-dependent and mainly useful for PyTorch-based inspection or continued experimentation.
- `config.yaml`: resolved Hydra training config.
- `metrics.csv`: training metrics from the uploaded checkpoint run.
- `metrics_summary.csv`: compact 5-seed final evaluation summary.
- `metadata.json`: compact metadata for inference and provenance.

## Limitations

This compact CNN is near MNIST saturation. Remaining errors are expected to be rare and often visually ambiguous or unusually written digits.