Here are important things to note for effective appication of MI-NiDIA Software.
- The software has only been validated with varying floc length sizes under different flocculation conditions (data from Non-Intrusive Dynamic Image Analysis), so data input should be floc data containing either a group or multiple groups of floc count per Tf.
- Your input does not need to be scaled as the software will upscale your dataset, as such, a column for your original timestamp (Tf) should be included and labelled as Tf in the datasheet.
- Users are encourage to downgrade their Tensorflow and Keras version (to 2.12.0 to 2.15.0) in a situation where the commands used in this code are no longer compatible with the latest version of Tensorflow.
- In an instance where the software is to be run on a system with lesser GPU power, the n_job is advised to be set as 1 (i.e. n_job = 1) in the minidiaModel.py file.