Major:
- [Allow to train with large Zarr/H5 files without loading data in memory](https://github.com/BiaPyX/BiaPy/commit/009b4e2b4da15ecce2cb99cab36b0a9508adc073)
- [Adapt by_chunks to multichannel data](https://github.com/BiaPyX/BiaPy/commit/de38df853856caa2ffab5f5ae473cc0374d7b7d8)
- [Add option to reuse predictions in test](https://github.com/BiaPyX/BiaPy/commit/923ddad8aec028bc3362c4f0fc16a29c663a525c)
Minor:
- Calculate detection metrics in by_chunks setting
- [Add DATA.NORMALIZATION.CUSTOM_MODE option](https://github.com/BiaPyX/BiaPy/commit/cc62199b7c98872366e065097919f2fc77666829)
- [Add percentile normalization](https://github.com/BiaPyX/BiaPy/commit/e923efcb8b2497d5f8e36f2402d7be7794bf01d5)
Fix:
- [Correct bug when discading samples from training when using TRAIN.MINIMUM_FOREGROUND_PER](https://github.com/BiaPyX/BiaPy/commit/50eef618f5579c9669e03d2909290817afb591ff)
- [Fix PSNR calculation in I2I and SR workflows](https://github.com/BiaPyX/BiaPy/commit/6ab226aabbab57def0c24646b890e170d16c2f5a)
- Update notebooks and correct errors related to gdown package
- [Correct validation paths in generators when in memory is False](https://github.com/BiaPyX/BiaPy/commit/6fb537b1b7acee45934405f51c28a1bf209d7d23)
**Full Changelog**: https://github.com/BiaPyX/BiaPy/compare/v3.3.14...v3.4.0