Deepblink

Overview

This pre-release contains some of the following updates:

* A completely overhauled CLI
* More succinct models
* Finalized benchmarking scripts
* Various bug fixes and improved documentation

---

Authors
* BBQuercus
* zhanyinx

Overview

This release mainly focussed on the addition of benchmarking and some bug fixes.
* Benchmarking scripts for TrackMate, SpotLearn, DetNet, and deepBlink
* According changes in `pink.metrics`
* Intensity calculation as option in CLI

Note: Changelog was not updated for the release. Updates will be added in the next commit.

---

Authors
* BBQuercus
* zhanyinx

v0.0.5x
We're on zenodo!

This is a intermediate release without many feature changes to link zenodo and get a DOI for this repository.

Overview

> “It works fine on MY computer”

Those days are over as it's testing time. While the test-suite is not fully complete it gives us some more confidence over our commits.
Additional changes in this release include:
* Change from prediction matrix to coordinate list as default dataset component. Accompanying changes include the use of cell_size in the `dataset` class and a dynamic Resnet architecture.
* Scipy requirement for a new `deepblink.metrics.f1_cutoff_score` way of calculating the F1 score directly from coordinates.
* Depreciation of `deepblink.metrics.compute_score` and `deepblink.metrics.weighted_f1_coordinates`.
* Addition of `deepblink.optimizers.amsgrad` function.
* Movement of functions:
* `train_valid_split` to `util`
* `load_image` to `io`

The next release will include benchmarking scripts and is most likely the last release before the first minor `0.1.0` and non-pre-release.

---

Authors
* BBQuercus
* zhanyinx

Overview

Let's train and inference!
This new release sees the following features:
* Addition of command line interface (currently without verbosity capabilities).
* Bin/ folder with ready to use training script.
* Removal of src/ directory. Deepblink is now sailing GitHub on its own.
* Major docs overhaul with the addition of more detailed input / functionality descriptions and the addition of a usage overview and example in README.
* YinXiu Zhan is now listed as co-author on pypi.

---

Authors
* Bastian Eichenberger
* YinXiu Zhan

Overview

We're happy to have our first custom code release. All utility functions and training loop have been included. The CLI is still cookiecutter and not functional. v0.0.3 marks the first of five patches leading up to the first minor version:
* v0.0.3 Utility functions.
* v0.0.4 Command line interface for inference and script folder for training.
* v0.0.5 Evaluation added to measure performance on test dataset.
* v0.0.6 Addition of tests and completion of current TODO’s.
* V0.0.7 Benchmarks/functions added to compare deepblink to other methods.

---

Authors
* Bastian Eichenberger
* YinXiu Zhan