From bb483515e3ef04da1c8edb869cab3825cd2a152a Mon Sep 17 00:00:00 2001 From: Andrew Wang Date: Tue, 28 Jul 2026 17:34:44 +0100 Subject: [PATCH] Update README.md --- README.md | 22 ++++++++++++---------- 1 file changed, 12 insertions(+), 10 deletions(-) diff --git a/README.md b/README.md index b4580d4..8241e93 100644 --- a/README.md +++ b/README.md @@ -3,17 +3,17 @@ This repository contains a set benchmarks for evaluating the performance of different image reconstruction methods implemented in the DeepInverse library. -- **DeepInverse documentation**: https://deepinv.github.io +- **DeepInverse documentation**: https://deepinv.org - **DeepInverse repository**: https://github.com/deepinv/deepinv -**Leaderboards** are automatically generated and can be found in the [DeepInverse benchmarks documentation](https://deepinv.github.io/deepinv/benchmarks.html). +**Leaderboards** are automatically generated and can be found in the [DeepInverse benchmarks documentation](https://deepinv.org/deepinv/benchmarks.html). Benchmark results are stored in a HuggingFace repository: https://huggingface.co/datasets/deepinv/benchmarks/tree/main ### Evaluating Your Reconstruction Methods To evaluate your own reconstruction methods on these benchmarks, -install deepinv_bench: +install `deepinv_bench`: ```bash pip install git+https://github.com/deepinv/benchmarks.git#egg=deepinv_bench @@ -34,10 +34,10 @@ my_solver = dinv.models.RAM() # replace with your reconstruction method results = run_benchmark(my_solver, "benchmark_name") ``` -where `benchmark_name` is the name of the benchmark and `my_solver` is your reconstruction method which receives `(y, physics)` where +where `benchmark_name` is the name of the benchmark and `my_solver` is your out-of-the-box reconstruction method which receives `(y, physics)` where - `y` is a `torch.Tensor` containing the measurements, -- `physics` is the forward operator, see more details in the [DeepInverse physics documentation](https://deepinv.github.io/deepinv/user_guide/physics/intro.html). +- `physics` is the forward operator, see more details in the [DeepInverse physics documentation](https://deepinv.org/deepinv/user_guide/physics/intro.html). ### Adding New Solvers @@ -45,6 +45,8 @@ To add a new solver to an existing benchmark, open a new pull request on this re in the corresponding benchmark folder. Follow the structure of the existing solver files. The new solver will be automatically run once the pull request is merged, and the results will be added to the leaderboard. +Also consider adding your solver to the collection of [existing solvers in DeepInverse](https://deepinv.org/user_guide/reconstruction/pretrained-models.html)! + ### Adding New Benchmarks To create a new benchmark, open a new pull request adding a new folder following @@ -52,9 +54,9 @@ the structure given in the existing [benchmark_template](https://github.com/deep A new benchmark requires: -- A **dataset** from the [DeepInverse datasets](https://deepinv.github.io/deepinv/user_guide/training/datasets.html). -- A **forward operator** from the [DeepInverse operators](https://deepinv.github.io/deepinv/user_guide/physics/physics.html). -- A set of **reconstruction methods** from the [DeepInverse reconstructors](https://deepinv.github.io/deepinv/user_guide/reconstruction/introduction.html) and potentially other custom solvers. -- A set of **metrics** from the [DeepInverse metrics](https://deepinv.github.io/deepinv/user_guide/training/metric.html). +- A **dataset** from the [DeepInverse datasets](https://deepinv.org/deepinv/user_guide/training/datasets.html). +- A **forward operator** from the [DeepInverse operators](https://deepinv.org/deepinv/user_guide/physics/physics.html). +- A set of out-of-the-box **reconstruction methods** from the [DeepInverse reconstructors](https://deepinv.org/deepinv/user_guide/reconstruction/introduction.html) and potentially other custom solvers. +- A set of **metrics** from the [DeepInverse metrics](https://deepinv.org/deepinv/user_guide/training/metric.html). -If you would like to propose a new dataset, metric or forward operator, please [open an issue](https://github.com/deepinv/benchmarks/issues/new/choose). \ No newline at end of file +If you would like to propose a new dataset, metric or forward operator, please [open an issue](https://github.com/deepinv/benchmarks/issues/new/choose).