update

.gitattributes

@@ -46,6 +46,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.ogg filter=lfs diff=lfs merge=lfs -text
 *.wav filter=lfs diff=lfs merge=lfs -text
 # Image files - uncompressed
+*.svg filter=lfs diff=lfs merge=lfs -text
 *.bmp filter=lfs diff=lfs merge=lfs -text
 *.gif filter=lfs diff=lfs merge=lfs -text
 *.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,35 @@
+---
+license: apache-2.0
+---
+# AutoImagine-Puzzle Dataset Card
+
+## Dataset Details
+This dataset contains 22 image-based jigsaw puzzles created from real-world photographs. Each puzzle consists of irregular puzzle pieces that need to be assembled into the complete image.
+
+## Dataset Structure
+- `dataset` folder
+  - `00000` ~ `00010` folder
+    - Contains several puzzle pieces and their final positions. `0.png` is the puzzle with pieces removed, while `1.png` ~ `4.png` are **four** individual puzzle pieces. `correct.json` contains the ground truth positions of the puzzle pieces.
+  - `00011` ~ `00021` folder
+    - Contains several puzzle pieces and their final positions. `0.png` is the puzzle with pieces removed, while `1.png` ~ `6.png` are **six** individual puzzle pieces. `correct.json` contains the ground truth positions of the puzzle pieces.
+  - `src` folder
+    - Contains the source files for generating puzzle datasets. Each dataset (e.g., `00000`, `00001`) has a corresponding pair of files:
+      - `{dataset}.png`: Original image file that provides the visual content for the puzzle
+      - `{dataset}.svg`: SVG file defining the boundary paths for cutting the image into irregular puzzle pieces
+- `data` folder
+  - Contains the prompts used for the autonomous imagination algorithm;
+
+## Resources for More Information
+- Paper: https://arxiv.org/pdf/2411.18142
+- Code: https://github.com/future-item/autoimagine
+
+## Citation
+If you find our work helpful, please consider citing our paper.
+```
+@article{liu2024enhancing,
+  title={Enhancing visual reasoning with autonomous imagination in multimodal large language models},
+  author={Liu, Jingming and Li, Yumeng and Xiao, Boyuan and Jian, Yichang and Qin, Ziang and Shao, Tianjia and Ding, Yao-Xiang and Zhou, Kun},
+  journal={arXiv preprint arXiv:2411.18142},
+  year={2024}
+}
+```

data/prompt-placer

@@ -0,0 +1,19 @@
+You are a master of puzzles. Now you are given an image of a jigsaw puzzle and please help me finish it!
+
+In this task, you will receive a picture of an unfinished puzzle. The picture contains a nearly complete puzzle and 1 piece that have not yet been put into the whole.
+
+You have selected a puzzle piece, indicated by a red outline. Now you need to move it into place so it fits into the nearly complete puzzle. 
+You should think step by step as follows.
+1. Describe the contents of the highlighted puzzle piece.
+2. Based on the analysis in step 1, point out the missing parts of the nearly completed puzzle and infer the contents of the missing parts based on the whole puzzle.
+3. The highlighted piece is surrounded by at most four letters 'a', 'b', 'c' and 'd'. Based on the analysis above, you should choose the letter that is closest to the target position so that the highlighted piece can best fill in a missing black piece of the whole puzzle.
+
+Since the image is edited, please ignore possible artifacts. The black background does not contain any puzzle-related content, please ignore it.
+
+You should output the quoted letters in the last line, that is, ['a'], ['b'], ['c'] or ['d'].
+
+
+
+
+
+

data/prompt-searcher

@@ -0,0 +1,11 @@
+In this task, you will receive a picture of an unfinished puzzle. The picture contains a nearly complete puzzle and several pieces that have not yet been put into the whole.
+
+You should think step by step as follows.
+
+1. Which of these individual puzzle pieces do you think fits best into the whole? Describe the contents and location of the puzzle pieces because letters on the image are NOT their labels!
+
+2. There are four letters 'a', 'b', 'c' and 'd' painted in the image. Which letter is closest to the individual puzzle piece selected in step 1?  
+
+Since the image is edited, please ignore possible artifacts. The black background does not contain any puzzle-related content, please ignore it.
+
+You should output the quoted letters in the last line, that is, ['a'], ['b'], ['c'] or ['d'].

data/prompt-searcher-checking-dot

@@ -0,0 +1,10 @@
+In this task, you will receive a picture of an unfinished puzzle. The picture contains a nearly complete puzzle and several pieces that have not yet been put into the whole.
+
+There is a small red circle in the picture. What you need to do is check if the red circle falls on one of the single puzzle pieces.
+
+Note that you should output 'no' if the red circle falls on the nearly complete puzzle, which contains more than one piece.
+If the circle falls on the edge of the object and you find it difficult to judge, you should output ['no'] as well.
+
+Since the image is edited, please ignore possible artifacts. The black background does not contain any puzzle-related content, please ignore it.
+
+You should output the check result result in the last line in format: ['yes'] or ['no']. 

data/prompt-searcher-checking-mask

@@ -0,0 +1,12 @@
+In this task, you need to check if there is a single puzzle piece in the image.
+
+You should check one by one in the following order.
+
+1. If the image is pure black and contains nothing, you should output 'no' and skip the following checks.
+2. Otherwise, if the image shows a partially completed puzzle that contains multiple pieces, you should output 'no' and skip the following checks.
+3. Otherwise, if the image contains only one puzzle piece, please output 'yes' and skip the following checks.
+4. If the above points are not met, please output 'no'.
+
+Since the image is edited, please ignore possible artifacts. The black background does not contain any puzzle-related content, please ignore it.
+
+You should output the final conclusion on the last line in format: ['yes'] or ['no']. 

dataset/00000/correct.json

@@ -0,0 +1,26 @@
+[
+    {
+        "id": 1,
+        "image": "1.png",
+        "gt_x": 350,
+        "gt_y": 69
+    },
+    {
+        "id": 2,
+        "image": "2.png",
+        "gt_x": 448,
+        "gt_y": 86
+    },
+    {
+        "id": 3,
+        "image": "3.png",
+        "gt_x": 338,
+        "gt_y": 201
+    },
+    {
+        "id": 4,
+        "image": "4.png",
+        "gt_x": 451,
+        "gt_y": 201
+    }
+]

dataset/00001/correct.json

@@ -0,0 +1,26 @@
+[
+    {
+        "id": 1,
+        "image": "1.png",
+        "gt_x": 150,
+        "gt_y": 185
+    },
+    {
+        "id": 2,
+        "image": "2.png",
+        "gt_x": 250,
+        "gt_y": 188
+    },
+    {
+        "id": 3,
+        "image": "3.png",
+        "gt_x": 150,
+        "gt_y": 318
+    },
+    {
+        "id": 4,
+        "image": "4.png",
+        "gt_x": 259,
+        "gt_y": 317
+    }
+]

dataset/00002/correct.json

@@ -0,0 +1,26 @@
+[
+    {
+        "id": 1,
+        "image": "1.png",
+        "gt_x": 63,
+        "gt_y": 212
+    },
+    {
+        "id": 2,
+        "image": "2.png",
+        "gt_x": 159,
+        "gt_y": 200
+    },
+    {
+        "id": 3,
+        "image": "3.png",
+        "gt_x": 51,
+        "gt_y": 329
+    },
+    {
+        "id": 4,
+        "image": "4.png",
+        "gt_x": 148,
+        "gt_y": 315
+    }
+]

dataset/00003/correct.json

@@ -0,0 +1,26 @@
+[
+    {
+        "id": 1,
+        "image": "1.png",
+        "gt_x": 61,
+        "gt_y": 197
+    },
+    {
+        "id": 2,
+        "image": "2.png",
+        "gt_x": 159,
+        "gt_y": 215
+    },
+    {
+        "id": 3,
+        "image": "3.png",
+        "gt_x": 249,
+        "gt_y": 186
+    },
+    {
+        "id": 4,
+        "image": "4.png",
+        "gt_x": 351,
+        "gt_y": 186
+    }
+]

dataset/00004/correct.json

@@ -0,0 +1,26 @@
+[
+    {
+        "id": 1,
+        "image": "1.png",
+        "gt_x": 315,
+        "gt_y": 202
+    },
+    {
+        "id": 2,
+        "image": "2.png",
+        "gt_x": 420,
+        "gt_y": 188
+    },
+    {
+        "id": 3,
+        "image": "3.png",
+        "gt_x": 286,
+        "gt_y": 332
+    },
+    {
+        "id": 4,
+        "image": "4.png",
+        "gt_x": 407,
+        "gt_y": 318
+    }
+]

dataset/00005/correct.json

@@ -0,0 +1,26 @@
+[
+    {
+        "id": 1,
+        "image": "1.png",
+        "gt_x": 63,
+        "gt_y": 70
+    },
+    {
+        "id": 2,
+        "image": "2.png",
+        "gt_x": 180,
+        "gt_y": 68
+    },
+    {
+        "id": 3,
+        "image": "3.png",
+        "gt_x": 312,
+        "gt_y": 81
+    },
+    {
+        "id": 4,
+        "image": "4.png",
+        "gt_x": 311,
+        "gt_y": 197
+    }
+]

dataset/00006/correct.json

@@ -0,0 +1,26 @@
+[
+    {
+        "id": 1,
+        "image": "1.png",
+        "gt_x": 165,
+        "gt_y": 85
+    },
+    {
+        "id": 2,
+        "image": "2.png",
+        "gt_x": 418,
+        "gt_y": 84
+    },
+    {
+        "id": 3,
+        "image": "3.png",
+        "gt_x": 522,
+        "gt_y": 68
+    },
+    {
+        "id": 4,
+        "image": "4.png",
+        "gt_x": 421,
+        "gt_y": 214
+    }
+]