InstruceBLIP

iconqa_data_preprocess.py

+import json
+
+with open('problems.json', 'r') as f:
+    data = json.load(f)
+
+train, test, val = [], [], []
+
+for key, value in data.items():
+    split = value["split"]
+    ques_type = value["ques_type"]
+    if ques_type == "choose_txt":
+        data = value
+        data['id'] = key
+        if split == "train" :
+            train.append(data)
+        elif split == "test":
+            test.append(data)
+        elif split == "val":
+            val.append(data)
+
+with open('/input/iconqa/annotations/train.json', 'w') as train_file:
+    json.dump(train, train_file, ensure_ascii=False, indent=4)
+
+with open('/input/iconqa/annotations/test.json', 'w') as test_file:
+    json.dump(test, test_file, ensure_ascii=False, indent=4)
+
+with open('/input/iconqa/annotations/val.json', 'w') as val_file:
+    json.dump(val, val_file, ensure_ascii=False, indent=4)
\ No newline at end of file

lavis/common/annotator/canny/__init__.py

+import cv2
+
+
+class CannyDetector:
+    def __call__(self, img, low_threshold, high_threshold):
+        return cv2.Canny(img, low_threshold, high_threshold)

lavis/common/annotator/ckpts/download.sh

+#! /bin/bash
+
+wget https://huggingface.co/lllyasviel/ControlNet/resolve/main/annotator/ckpts/dpt_hybrid-midas-501f0c75.pt
+wget https://huggingface.co/lllyasviel/ControlNet/resolve/main/annotator/ckpts/network-bsds500.pth
+