Transition away from Conda

[mrl280]

Description:
Currently, we use Conda for this example because it is the easiest way to install GDAL, which is required to read the spectrogram masks from the Spectrogram Sensing dataset.

However, Google Colab doesn't provide native Conda support. Instead, we'd have to use condacolab, which at the time of writing only supports Python 3.10 and older versions.

Consequently, we'd like to transition this example away from Conda, which requires converting the Spectrogram Sensing dataset so we can remove GDAL as a project dependency. Luckily, we are converting the MathWork's Spectorgram Sensing dataset to an HDF5 RIA dataset anyway, as part of this issue here: Convert MathWorks' Spectrogram Sensing dataset to HDF5.

Once this transition is complete, this project needs to be updated to download and use the new HDF5 file, instead of the individual .PNG and .HDF frames.

Project Area:

• Dataset
• Virtual environment

Additional context:
As part of these changes, we should consider downgrading to Python 3.10.

We do not want to use the RIA radio dataset or its dataset builder; only the source file. The rationale is that in the future, we plan to create another example that leverages RIA tooling to simplify the end-to-end model training and validation process. Therefore, we aim to limit this example as much as possible to what's achievable with the current tooling.

We will need to replace the environment.yaml file with a requirements.txt file containing the new environment definition.