B-roll storage system

Description

System for storing b-roll images and video in an easily accessible manner Uses Express for backend, MongoDB for storage, RAM+ for tagging media

How to start the backend (funky AI stuff)

git clone https://github.com/sitechtimes/b-roll-storage
cd b-roll-storage
npm i
git clone https://github.com/xinyu1205/recognize-anything
python -m venv .venv

if on windows (I think): .\.venv\Scripts\pip.exe install -r recognize-anything/requirements.txt or if on linux: ./.venv/bin/pip install -r recognize-anything/requirements.txt

To actually start it, npm run start (don't forget the .env)

Now there are two weird errors that will occur from here when you attempt to get the AI tagging model to run:

The first of which is ImportError: cannot import name 'apply_chunking_to_forward' from 'transformers.modeling_utils' to which you need to edit /recognise-everything/ram/models/bert.py and change

from transformers.modeling_utils import (
    PreTrainedModel,
    apply_chunking_to_forward,
    find_pruneable_heads_and_indices,
    prune_linear_layer,
)

to

from transformers.modeling_utils import (
    PreTrainedModel,
)

from transformers.pytorch_utils import (
    apply_chunking_to_forward,
    find_pruneable_heads_and_indices,
    prune_linear_layer,
)

The second error is ImportError: cannot import name 'find_pruneable_heads_and_indices' from 'transformers.pytorch_utils' to which you need to run .venv/bin/pip install "transformers==4.57.1" (this happens because it installs the incorrect version from the requirements.txt, so you can change that instead).