Hf implementation

.gitignore

@@ -1,2 +1,6 @@
 scripts/*.sh
 **/__pycache__
+*events.out*
+output/**/*.json
+*.png
+.ipynb_checkpoints

README.md

@@ -1,70 +1,95 @@
-# RxnScribe 
+# MolDetect and MolCoref
 
-This is the repository for RxnScribe, a sequence generation model for reaction diagram parsing.
-Try our [demo](https://huggingface.co/spaces/yujieq/RxnScribe) on Hugging Face!
-
-![](assets/model.png)
+This is the repository for MolDetect and MolCoref, two sequence generation models for reaction diagram parsing.
 
 ## Quick Start
 Run the following command to install the package and its dependencies:
 ```
-git clone git@github.com:thomas0809/RxnScribe.git
+git clone git@github.com:Ozymandias314/MolDetect.git
 cd RxnScribe
 python setup.py install
 ```
 
-Download the checkpoint and use RxnScribe to extract reactions from a diagram:
-```python
+Download the checkpoint and use MolDetect to extract molecules from reaction diagrams:
+
+```python 
 import torch
-from rxnscribe import RxnScribe
+from rxnscribe import MolDetect
 from huggingface_hub import hf_hub_download
 
-ckpt_path = hf_hub_download("yujieq/RxnScribe", "pix2seq_reaction_full.ckpt")
-model = RxnScribe(ckpt_path, device=torch.device('cpu'))
+ckpt_path = hf_hub_download("Ozymandias314/MolDetectCkpt", "best.ckpt")
+model = MolDetect(ckpt_path, device=torch.device('cpu'))
 
 image_file = "assets/jacs.5b12989-Table-c3.png"
-predictions = model.predict_image_file(image_file, molscribe=True, ocr=True)
+predictions = model.predict_image_file(image_file)
 ```
 The predictions will be in the following format:
 ```python
 [
-    {  # First reaction
-        'reactants': [
-            {
-                'category': '[Mol]', 'category_id': 1, 'bbox': (0.1550, 0.0246, 0.2851, 0.2614),
-                'smiles': '*OC(=O)c1ccccc1C#Cc1ccccc1', 'molfile': '(omitted)' 
-            }, 
-            # ... more reactants 
-        ],
-        'conditions': [
-            {
-                'category': '[Txt]', 'category_id': 2, 'bbox': (0.2941, 0.0641, 0.3811, 0.1450),
-                'text': ['CIBcat', '(1.4 equiv)']
-            }, 
-            # ... more conditions
-        ],
-        'products': [ 
-            # ...
-        ]
+    {   #first bbox
+        'category': '[Sup]', 
+        'bbox': (0.0050025012506253125, 0.38273870663142223, 0.9934967483741871, 0.9450094869920168), 
+        'category_id': 4, 
+        'score': -0.07593922317028046
     },
-    # More reactions
+    #More bounding boxes
 ]
 ```
-We provide a function to visualize the prediction:
+We provide a function to visualize the predicted bboxes:
 ```python
-visualize_images = model.draw_predictions(predictions, image_file=image_file)
+visualize_images = model.draw_bboxes(predictions, image_file = image_file)
 ```
-Each predicted reaction will be visualized in a separate image, where 
-<b style="color:red">red boxes are <i><u style="color:red">reactants</u></i>,</b>
-<b style="color:green">green boxes are <i><u style="color:green">reaction conditions</u></i>,</b>
-<b style="color:blue">blue boxes are <i><u style="color:blue">products</u></i>.</b>
 
-<img src="assets/output/output0.png" width="384"/> <img src="assets/output/output1.png" width="384"/> 
+Each predicted diagram will be visualized in a seperate image, where
+<b style="color:red">red boxes are <i><u style="color:red">molecules</u></i>,</b>
+<b style="color:green">green boxes are <i><u style="color:green">text</u></i>,</b>
+<b style="color:blue">blue boxes are <i><u style="color:blue">identifiers</u></i>.</b> 
+<b style="color:gold">gold boxes are <i><u style="color:blue">supplementary information</u></i>.</b> 
 
-This [notebook](notebook/predict.ipynb) shows how to run RxnScribe and visualize the prediction.
+<img src="assets/output/output2.png" width = "384"/>
 
-For development or reproducing the experiments, follow the instructions below.
+To detect bounding boxes and also the coreference, download the coreference checkpoint for MolDetect on huggingface and also set the coref field to be true:
+
+```python 
+import torch
+from rxnscribe import MolDetect
+from huggingface_hub import hf_hub_download
+
+ckpt_path = hf_hub_download("Ozymandias314/MolDetectCkpt", "coref_best.ckpt")
+model = MolDetect(ckpt_path, device=torch.device('cpu'), coref = True)
+
+image_file = "assets/jacs.5b12989-Table-c3.png"
+predictions = model.predict_image_file(image_file, coref = True)
+```
 
+The predictions will be in the following format:
+
+```python
+{
+    'bboxes': [
+        {   #first bbox
+            'category': '[Sup]', 
+            'bbox': (0.0050025012506253125, 0.38273870663142223, 0.9934967483741871, 0.9450094869920168), 
+            'category_id': 4, 
+            'score': -0.07593922317028046
+        },
+        #More bounding boxes
+    ],
+    'coref': [
+        [0, 1],
+        [3, 4],
+        #More coref pairs
+    ]
+}
+```
+
+Similarly, we provide a function to visualize the predicted bboxes:
+
+```python
+visualize_images = model.draw_bboxes(predictions, image_file = image_file, coref = True)
+```
+
+For development or reproducing the experiments, follow the instructions below.
 ## Requirements
 Install the required packages
 ```
@@ -73,7 +98,7 @@ pip install -r requirements.txt
 
 ## Data
 Download the reaction diagrams from this [link](https://huggingface.co/yujieq/RxnScribe/blob/main/images.zip), 
-and save them to `data/parse/images/`.
+and save them to `data/detect/images/`.
 
 The ground truth files can be found at [`data/parse/splits/`](data/parse/splits/).
 

data/detect/.ipynb_checkpoints/sample no mol-checkpoint.ipynb

@@ -0,0 +1,6 @@
+{
+ "cells": [],
+ "metadata": {},
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

data/detect/splits/annotations/.ipynb_checkpoints/Untitled-checkpoint.ipynb

@@ -0,0 +1,6 @@
+{
+ "cells": [],
+ "metadata": {},
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

data/detect/splits/annotations/.ipynb_checkpoints/get rid of class 2 and 4-checkpoint.ipynb

@@ -0,0 +1,6 @@
+{
+ "cells": [],
+ "metadata": {},
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

data/detect/splits/annotations/.ipynb_checkpoints/split no mol-checkpoint.ipynb

@@ -0,0 +1,6 @@
+{
+ "cells": [],
+ "metadata": {},
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

data/detect/splits/annotations/Untitled.ipynb

@@ -0,0 +1,104 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "90feaae3-97b6-4516-aba0-89f91fdc8564",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "categories\n",
+      "licenses\n",
+      "images\n",
+      "annotations\n",
+      "categories\n",
+      "licenses\n",
+      "images\n",
+      "annotations\n"
+     ]
+    }
+   ],
+   "source": [
+    "import json\n",
+    "\n",
+    "# List the names of the JSON files to be read\n",
+    "files = ['detect_test.json', 'detect_train.json', 'detect_val.json']\n",
+    "\n",
+    "# Create an empty dictionary to hold the combined data\n",
+    "combined_data = {}\n",
+    "\n",
+    "# Loop over each file\n",
+    "for filename in files:\n",
+    "    # Open the file and load its contents into a dictionary\n",
+    "    with open(filename, 'r') as f:\n",
+    "        data = json.load(f)\n",
+    "        \n",
+    "    # Combine the data into the combined_data dictionary\n",
+    "    for key, value in data.items():\n",
+    "        if key in combined_data and key != \"info\":\n",
+    "            print(key)\n",
+    "            combined_data[key].extend(value)\n",
+    "        else:\n",
+    "            combined_data[key] = value\n",
+    "\n",
+    "# Write the combined data to a new JSON file\n",
+    "with open('detect_combined.json', 'w') as f:\n",
+    "    json.dump(combined_data, f)\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "ef9b649b-00e3-4e4e-815d-345fa0eac053",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "[{'id': 1, 'width': 3396, 'height': 1776, 'file_name': 'jacs.6b00143-Table-c2.png', 'license': 0, 'flickr_url': '', 'coco_url': '', 'date_captured': 0, 'bboxes': [{'bbox': [2390.35, 60.38, 146.74, 112.34], 'category_id': 1, 'id': 5}, {'bbox': [1262.95, 198.63, 49.38, 36.4], 'category_id': 3, 'id': 6}, {'bbox': [878.06, 201.87, 81.18, 34.45], 'category_id': 3, 'id': 7}, {'bbox': [2443.58, 201.87, 28.61, 35.1], 'category_id': 3, 'id': 8}, {'bbox': [289.0, 259.21, 2816.1, 1237.18], 'category_id': 4, 'id': 9}, {'bbox': [1486.22, 2.61, 310.95, 118.83], 'category_id': 2, 'id': 10}, {'bbox': [1908.11, 9.1, 357.68, 188.28], 'category_id': 1, 'id': 11}, {'bbox': [1127.7, 41.51, 272.44, 171.58], 'category_id': 1, 'id': 12}, {'bbox': [848.15, 59.84, 146.35, 109.71], 'category_id': 1, 'id': 13}, {'bbox': [1483.62, 148.0, 319.39, 42.89], 'category_id': 2, 'id': 14}, {'bbox': [2043.11, 201.87, 27.96, 35.1], 'category_id': 3, 'id': 15}], 'caption': 'Table 2. Substrate Scope for Kinetic Resolutions of Sulfoximines and a Sulfondiimide with Enal 2ba', 'pdf': {'Page': 2, 'DPI': 100, 'Width': 869, 'Height': 1137, 'CaptionBB': [82, 656, 648, 677], 'ImageBB': [10, 683, 859, 1127]}}, {'id': 2, 'width': 1356, 'height': 688, 'file_name': 'jacs.5b12989-Table-c3.png', 'license': 0, 'flickr_url': '', 'coco_url': '', 'date_captured': 0, 'bboxes': [{'bbox': [6.65, 262.6, 1338.95, 417.44], 'category_id': 4, 'id': 16}, {'bbox': [965.41, 10.46, 176.88, 169.79], 'category_id': 1, 'id': 17}, {'bbox': [624.51, 10.46, 176.0, 169.79], 'category_id': 1, 'id': 18}, {'bbox': [209.94, 14.9, 176.88, 165.3], 'category_id': 1, 'id': 19}, {'bbox': [399.03, 42.42, 119.18, 58.82], 'category_id': 2, 'id': 20}, {'bbox': [399.03, 122.3, 172.44, 62.38], 'category_id': 2, 'id': 21}, {'bbox': [815.38, 103.68, 134.27, 33.07], 'category_id': 2, 'id': 22}, {'bbox': [288.95, 206.65, 23.3, 29.52], 'category_id': 3, 'id': 23}, {'bbox': [707.96, 206.65, 25.97, 29.52], 'category_id': 3, 'id': 24}, {'bbox': [1051.52, 206.65, 49.94, 29.52], 'category_id': 3, 'id': 25}, {'bbox': [815.38, 35.32, 100.54, 59.7], 'category_id': 2, 'id': 26}], 'caption': 'Table 3. Mechanistic Insight from O-Alkyl Group Variance of the Oxyboration Reaction ', 'pdf': {'Page': 3, 'DPI': 100, 'Width': 869, 'Height': 1137, 'CaptionBB': [449, 91, 784, 125], 'ImageBB': [448, 131, 787, 303]}}, {'id': 12, 'width': 1344, 'height': 800, 'file_name': 'ja9810742-Table-c3.png', 'license': 0, 'flickr_url': '', 'coco_url': '', 'date_captured': 0, 'bboxes': [{'bbox': [612.17, 106.36, 169.22, 43.29], 'category_id': 2, 'id': 176}, {'bbox': [619.39, 27.91, 155.81, 45.35], 'category_id': 2, 'id': 177}, {'bbox': [806.23, 2.1, 277.62, 183.68], 'category_id': 1, 'id': 178}, {'bbox': [364.43, 10.36, 235.29, 171.29], 'category_id': 1, 'id': 179}, {'bbox': [935.26, 184.81, 37.1, 37.1], 'category_id': 3, 'id': 180}, {'bbox': [467.65, 185.84, 37.1, 40.2], 'category_id': 3, 'id': 181}, {'bbox': [244.68, 75.39, 52.58, 41.23], 'category_id': 3, 'id': 182}, {'bbox': [2.1, 310.75, 1337.75, 483.03], 'category_id': 4, 'id': 183}, {'bbox': [811.39, 232.3, 281.74, 44.32], 'category_id': 2, 'id': 184}], 'caption': 'Table 3. Allylation of 3-Phenylpropanal 1a with Homoallylic Alcohols 2a ', 'pdf': {'Page': 2, 'DPI': 100, 'Width': 850, 'Height': 1100, 'CaptionBB': [74, 544, 385, 569], 'ImageBB': [74, 570, 410, 770]}}, {'id': 17, 'width': 1168, 'height': 824, 'file_name': 'ja960062i-Figure-c4.png', 'license': 0, 'flickr_url': '', 'coco_url': '', 'date_captured': 0, 'bboxes': [{'bbox': [51.41, 500.95, 333.73, 110.45], 'category_id': 1, 'id': 208}, {'bbox': [421.42, 220.25, 159.35, 34.96], 'category_id': 2, 'id': 209}, {'bbox': [451.19, 429.71, 655.88, 385.82], 'category_id': 1, 'id': 210}, {'bbox': [653.2, 5.48, 507.03, 351.8], 'category_id': 1, 'id': 211}, {'bbox': [5.7, 15.05, 334.78, 341.17], 'category_id': 1, 'id': 212}, {'bbox': [364.0, 119.25, 266.75, 36.02], 'category_id': 2, 'id': 213}], 'caption': 'Figure 4. Scheme of synthesis of oligonucleotide-CHEVP conjugate by the diamine method. ', 'pdf': {'Page': 4, 'DPI': 100, 'Width': 850, 'Height': 1100, 'CaptionBB': [73, 260, 408, 286], 'ImageBB': [95, 50, 387, 256]}}, {'id': 23, 'width': 1352, 'height': 1080, 'file_name': 'ja905415r-Table-c5.png', 'license': 0, 'flickr_url': '', 'coco_url': '', 'date_captured': 0, 'bboxes': [{'bbox': [431.53, 84.4, 251.45, 105.12], 'category_id': 2, 'id': 258}, {'bbox': [990.35, 31.45, 279.31, 258.41], 'category_id': 1, 'id': 259}, {'bbox': [710.24, 4.97, 254.23, 291.86], 'category_id': 1, 'id': 260}, {'bbox': [76.18, 59.32, 336.45, 234.72], 'category_id': 1, 'id': 261}, {'bbox': [2.32, 321.3, 1343.99, 753.13], 'category_id': 4, 'id': 262}, {'bbox': [1178.47, 265.56, 56.35, 42.41], 'category_id': 3, 'id': 263}, {'bbox': [306.11, 266.96, 54.96, 41.01], 'category_id': 3, 'id': 264}, {'bbox': [431.53, 204.25, 248.66, 49.38], 'category_id': 2, 'id': 265}, {'bbox': [863.53, 271.14, 53.56, 42.41], 'category_id': 3, 'id': 266}], 'caption': 'Table 5. Comparison of the enantioselectivities of 2a and 3a, isolated from the reaction of 1a using different Au catalysts.a ', 'pdf': {'Page': 7, 'DPI': 100, 'Width': 818, 'Height': 1088, 'CaptionBB': [58, 57, 365, 81], 'ImageBB': [58, 89, 396, 359]}}, {'id': 27, 'width': 1348, 'height': 1132, 'file_name': 'ja9039289-Table-c3.png', 'license': 0, 'flickr_url': '', 'coco_url': '', 'date_captured': 0, 'bboxes': [{'bbox': [949.48, 205.43, 374.46, 47.28], 'category_id': 2, 'id': 290}, {'bbox': [1110.15, 253.63, 39.98, 41.47], 'category_id': 3, 'id': 291}, {'bbox': [1110.15, 148.46, 38.52, 42.9], 'category_id': 3, 'id': 292}, {'bbox': [743.53, 253.63, 35.59, 42.9], 'category_id': 3, 'id': 293}, {'bbox': [281.96, 222.95, 301.44, 45.82], 'category_id': 2, 'id': 294}, {'bbox': [636.9, 142.62, 254.69, 108.63], 'category_id': 1, 'id': 295}, {'bbox': [1050.26, 95.88, 155.37, 48.74], 'category_id': 2, 'id': 296}, {'bbox': [21.97, 139.7, 206.49, 107.16], 'category_id': 1, 'id': 297}, {'bbox': [2.98, 303.29, 1341.41, 827.21], 'category_id': 4, 'id': 298}, {'bbox': [298.03, 3.86, 269.3, 180.2], 'category_id': 2, 'id': 299}], 'caption': 'Table 3. Fluoride Effect on Ni- and Co-Catalyzed Biaryl Cross-Coupling in the Presence of Phosphine Ligands ', 'pdf': {'Page': 5, 'DPI': 100, 'Width': 818, 'Height': 1088, 'CaptionBB': [425, 57, 703, 81], 'ImageBB': [425, 89, 762, 372]}}, {'id': 49, 'width': 1352, 'height': 1820, 'file_name': 'ja408733f-Figure-c2.png', 'license': 0, 'flickr_url': '', 'coco_url': '', 'date_captured': 0, 'bboxes': [{'bbox': [364.33, 441.71, 98.0, 35.95], 'category_id': 2, 'id': 516}, {'bbox': [523.35, 309.95, 305.23, 242.6], 'category_id': 1, 'id': 517}, {'bbox': [856.55, 377.53, 113.92, 35.95], 'category_id': 2, 'id': 518}, {'bbox': [572.78, 216.29, 58.22, 34.88], 'category_id': 2, 'id': 519}, {'bbox': [672.5, 216.82, 83.15, 34.88], 'category_id': 2, 'id': 520}, {'bbox': [388.72, 367.98, 73.08, 36.48], 'category_id': 2, 'id': 521}, {'bbox': [522.59, 689.56, 344.18, 377.03], 'category_id': 1, 'id': 522}, {'bbox': [690.63, 565.06, 167.74, 118.86], 'category_id': 1, 'id': 523}, {'bbox': [460.72, 565.06, 167.75, 118.86], 'category_id': 1, 'id': 524}, {'bbox': [25.36, 225.93, 305.23, 327.38], 'category_id': 1, 'id': 525}, {'bbox': [1015.24, 206.07, 305.23, 346.48], 'category_id': 1, 'id': 526}, {'bbox': [391.22, 2.9, 519.85, 152.47], 'category_id': 1, 'id': 527}, {'bbox': [29.53, 1144.91, 1292.75, 668.44], 'category_id': 1, 'id': 528}], 'caption': 'Figure 2. Molecular structure of [κ4-Tptm]ZnF·H2O.', 'pdf': {'Page': 2, 'DPI': 100, 'Width': 869, 'Height': 1137, 'CaptionBB': [82, 578, 338, 595], 'ImageBB': [82, 116, 420, 571]}}, {'id': 70, 'width': 1344, 'height': 1332, 'file_name': 'ja208286b-Table-c1.png', 'license': 0, 'flickr_url': '', 'coco_url': '', 'date_captured': 0, 'bboxes': [{'bbox': [523.14, 256.74, 247.36, 173.92], 'category_id': 2, 'id': 765}, {'bbox': [4.02, 627.29, 1337.34, 307.46], 'category_id': 4, 'id': 766}, {'bbox': [207.66, 355.22, 27.04, 33.71], 'category_id': 3, 'id': 767}, {'bbox': [968.8, 71.46, 270.74, 504.42], 'category_id': 1, 'id': 768}, {'bbox': [484.74, 3.02, 305.79, 195.63], 'category_id': 1, 'id': 769}, {'bbox': [105.84, 71.46, 255.72, 275.74], 'category_id': 1, 'id': 770}, {'bbox': [695.06, 156.58, 48.74, 37.06], 'category_id': 3, 'id': 771}, {'bbox': [1167.44, 355.22, 48.73, 38.72], 'category_id': 3, 'id': 772}], 'caption': 'Table 1. Optimization Studies for the Ruthenium-Catalyzed Meta Sulfonation of 2-Phenylpyridine (1)a ', 'pdf': {'Page': 2, 'DPI': 100, 'Width': 844, 'Height': 1112, 'CaptionBB': [71, 92, 404, 122], 'ImageBB': [71, 133, 407, 466]}}, {'id': 76, 'width': 1356, 'height': 1136, 'file_name': 'ja2014746-Figure-c1.png', 'license': 0, 'flickr_url': '', 'coco_url': '', 'date_captured': 0, 'bboxes': [{'bbox': [269.68, 430.42, 20.43, 28.16], 'category_id': 3, 'id': 838}, {'bbox': [37.04, 5.39, 527.1, 139.56], 'category_id': 1, 'id': 839}, {'bbox': [6.0, 631.97, 640.29, 481.94], 'category_id': 1, 'id': 840}, {'bbox': [668.32, 622.33, 680.23, 506.73], 'category_id': 1, 'id': 841}, {'bbox': [650.89, 32.01, 86.31, 37.5], 'category_id': 2, 'id': 842}, {'bbox': [603.56, 88.22, 176.54, 62.64], 'category_id': 2, 'id': 843}, {'bbox': [631.67, 287.91, 86.31, 36.02], 'category_id': 2, 'id': 844}, {'bbox': [22.25, 240.57, 518.23, 160.27], 'category_id': 1, 'id': 845}, {'bbox': [821.0, 5.39, 501.96, 135.12], 'category_id': 1, 'id': 846}, {'bbox': [806.21, 234.66, 504.91, 163.23], 'category_id': 1, 'id': 847}, {'bbox': [584.33, 344.11, 175.06, 62.65], 'category_id': 2, 'id': 848}, {'bbox': [288.5, 148.86, 16.79, 30.11], 'category_id': 3, 'id': 849}, {'bbox': [1047.81, 429.83, 20.43, 26.37], 'category_id': 3, 'id': 850}, {'bbox': [1066.24, 151.03, 20.43, 27.56], 'category_id': 3, 'id': 851}], 'caption': 'Figure 1. (Left) ORTEP drawing of 2 (50% probability level). One of the tert-butyl groups on Si3 is disordered, and the minor contribution (occupancy factor 0.299) and hydrogen atoms are omitted for clarity. # indicates the following symmetry transformation:  x,  y þ 1,  z. Selected bond lengths (Å) and angles (deg): Si1 Si2 = 2.3675(6), Si1 Si3 = 2.3737(6), Si1 C1 = 1.8174(14), C1 C2 = 1.423(2), C1 C3# = 1.422(2), C2 C3 = 1.3700(19). Si2 Si1 Si3 = 122.86(2), Si2 Si1 C1 = 117.52(5), Si3 Si1 C1 = 119.58(5), Si1 C1 C2 = 120.66(11), Si1 C1 C3# = 124.39(11), C2 C1 C3# = 114.94(12), C1 C2 C3 = 122.78(13), C2 C3 C1# = 122.29(13). (Right) ORTEP drawing of 4 (50% probability level). Hydrogen atoms are omitted for clarity. # indicates the following symmetry transformation:  x þ 1/2,  y þ 1/2, z. Selected bond lengths (Å) and angles (deg): Si1 Si2 = 2.3554(14), Si1 Si3 = 2.3524(14), Si1 C2 = 1.9108(19), C1 C2 = 1.387(3), C1 C2# = 1.387(3), C2 C3 = 1.395(5), C3 C4 = 1.386(4), C4 C3# = 1.386(4). Si2 Si1 Si3 = 122.21(5), Si2 Si1 C2 = 117.15(5), Si3 Si1 C2 = 119.29(5), C2 C1 C2# = 123.2(4), Si1 C2 C1 = 120.8(3), Si1 C2 C3 = 121.73(17), C1 C2 C3 = 117.5(2), C2 C3 C4 = 121.0(3), C3 C4 C3# = 119.7(4). ', 'pdf': {'Page': 2, 'DPI': 100, 'Width': 844, 'Height': 1112, 'CaptionBB': [71, 430, 405, 707], 'ImageBB': [71, 133, 410, 417]}}, {'id': 87, 'width': 1348, 'height': 736, 'file_name': 'ja063878k-Table-c2.png', 'license': 0, 'flickr_url': '', 'coco_url': '', 'date_captured': 0, 'bboxes': [{'bbox': [443.28, 145.42, 21.31, 30.46], 'category_id': 3, 'id': 939}, {'bbox': [630.33, 103.74, 154.49, 88.41], 'category_id': 2, 'id': 940}, {'bbox': [650.66, 38.67, 101.63, 31.48], 'category_id': 2, 'id': 941}, {'bbox': [302.99, 10.21, 229.71, 165.67], 'category_id': 1, 'id': 942}, {'bbox': [70.19, 4.11, 143.3, 169.74], 'category_id': 1, 'id': 943}, {'bbox': [8.18, 208.44, 1335.72, 517.46], 'category_id': 4, 'id': 944}, {'bbox': [853.98, 15.29, 320.19, 168.72], 'category_id': 1, 'id': 945}, {'bbox': [142.37, 146.43, 81.29, 36.57], 'category_id': 3, 'id': 946}], 'caption': 'Table 2. Alkenylation with 2-Substituted Vinylsilanes', 'pdf': {'Page': 1, 'DPI': 100, 'Width': 850, 'Height': 1100, 'CaptionBB': [440, 514, 705, 526], 'ImageBB': [439, 528, 776, 712]}}]\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(combined_data['images'][0:10])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "b26c4c01-9dd8-49a6-b8a7-cf43f8662c6c",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}