Clarisolve

A DL-based super-resolution and colorization tool.

Requirements

You must have Python with minimum version 3.10.0 installed your system.

To install the required packages, you can run:

pip install -r requirements.txt

Usage

Warning

This tool does not catch errors. As a result, running it incorrectly may produce incoherent traces.

Evaluation

A GUI tool is provided to super-resolve and colorize images. To run, use:

python main.py

CLI options are also present, see the super-resolve.py and colorize.py files for details.

Training

Note

For super-resolution models, a .h5 file is required for both datasets, and for colorization, a directory is required.

You can train any model yourself using train.py as follows:

python train.py --model { "srcnn", "srcnnc", "srres", "iccnn", "icres" } \
                --train-data TRAIN_DATA \
                --eval-data EVAL_DATA \
                --output-dir OUTPUT_DIR \
                [--checkpoint-path CHECKPOINT_PATH] \
                [--learn-rate LEARN_RATE] \
                [--end-epoch END_EPOCH] \
                [--num-workers NUM_WORKERS] \
                [--seed SEED]

Datasets

A utility script util/make.py is provided for .h5 file creation.