Multiple changes to VQA data prep (#21)

- Fixed the broken links for PathVQA, SLAKE, RadVQA (Github dataset
links for PathVQA and SLAKE were originally there and now have been
removed)
- The original PathVQA dataset could not be found except for a hugging
face link, this required adding additional processing steps to the data.
New script to process from parquet files have been added
- Readme was also updated
- Instruction tuning files have been removed as providing them would be
a form of re-distribution of data.

---------

Signed-off-by: Vishwesh Nath <[email protected]>

monai_vila2d/data_prepare/README.md

@@ -28,9 +28,9 @@ experts
 
 | Dataset   | QA/Text Pairs  | Images    | Link |
 |-----------|-----------|-----------|------|
-| PathVQA   | ~32,000   | ~4,000    | [PathVQA](https://github.com/UCSD-AI4H/PathVQA) |
-| RadVQA    | ~25,000   | ~7,000    | [RadVQA](https://github.com/abachaa/VQA-Med-2019) |
-| SLAKE     | ~45,000   | ~14,000   | [SLAKE](https://github.com/SLAKE-SLAKE/SLAKE) |
+| PathVQA   | ~32,000   | ~4,000    | [PathVQA](https://huggingface.co/datasets/flaviagiammarino/path-vqa) |
+| RadVQA    | ~25,000   | ~7,000    | [RadVQA](https://osf.io/89kps/) |
+| SLAKE     | ~45,000   | ~14,000   | [SLAKE](https://www.med-vqa.com/slake/) |
 | Medical-Diff-VQA | ~429,000  | 129,232  | [MIMIC-VQA](https://physionet.org/content/medical-diff-vqa/1.0.0) |
 | MIMIC-CXR-JPG | 270,784 | 270,784 | [MIMIC-CXR-JPG](https://physionet.org/content/mimic-cxr-jpg/2.1.0/) |
 | ChestXRay14 | 1,962 | 1,962 | [nih-chest-xray](https://cloud.google.com/healthcare-api/docs/resources/public-datasets/nih-chest#additional_labels) |

monai_vila2d/data_prepare/vqa/README.md

@@ -31,11 +31,11 @@ python slake_instruct_data_generate.py \
 Example command to run generate the instruction training data json file for PathVQA dataset:
 
 ```
-python pathvqa_instruct_data_generate.py \
-    --train_pkl /path/to/train_vqa.pkl \
-    --val_pkl /path/to/val_vqa.pkl \
-    --test_pkl /path/to/test_vqa.pkl \
-    --output_json /path/to/output/merged_pathvqa_instruct.json
+python pathvqa_instruction_gen_parquet.py --input_path /path/to/input/parquet/files --output_path /path/to/output/processed/dataset
+```
+Please make sure that the .csv files were succesfully generated from the prior command before running the next command
+```
+python pathvqa_instruction_generate.py --input_dir /path/to/output/processed/dataset --output_dir /path/to/output_directory
 ```
 
 ### MIMIC-VQA