toqi commited on
Commit
5789da0
·
verified ·
1 Parent(s): 678b340

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +216 -0
README.md ADDED
@@ -0,0 +1,216 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-nc-sa-4.0
3
+ task_categories:
4
+ - image-segmentation
5
+ - image-classification
6
+ tags:
7
+ - weed
8
+ - agriculture
9
+ - segmentation
10
+ - semantic-segmentation
11
+ - height-estimation
12
+ - regression
13
+ - growth-stage
14
+ - multi-task-learning
15
+ - weedsense
16
+ - precision-agriculture
17
+ - plant-phenotyping
18
+ - temporal
19
+ pretty_name: "WeedSense: Multi-Task Weed Analysis Dataset"
20
+ size_categories:
21
+ - 100K<n<1M
22
+ ---
23
+
24
+ # WeedSense Dataset
25
+
26
+ Multi-task temporal dataset of **16 weed species** for semantic segmentation, height regression, and growth stage classification from the **WeedSense** paper:
27
+
28
+ > **WeedSense: Multi-Task Learning for Weed Segmentation, Height Estimation, and Growth Stage Classification**
29
+ > Toqi Tahamid Sarker, Khaled R Ahmed, Taminul Islam, Cristiana Bernardi Rankrape, Karla Gage
30
+ > Southern Illinois University Carbondale, USA
31
+ > **ICCV 2025**
32
+ >
33
+ > [Paper (arXiv)](https://arxiv.org/abs/2508.14486) | [Project Page](https://weedsense.github.io/) | [Code](https://github.com/toqitahamid/WeedSense)
34
+
35
+ ## Overview
36
+
37
+ | Property | Value |
38
+ |---|---|
39
+ | Total frames | **120,341** |
40
+ | Weed species | **16** |
41
+ | Growth duration | **11 weeks** |
42
+ | Total videos | **349** |
43
+ | Frame resolution | **720 x 960 pixels** |
44
+ | Annotation types | **Segmentation masks, Height (cm), Growth stage (week)** |
45
+ | Semantic classes | **17** (16 species + background) |
46
+ | Height range | **0.2 - 155 cm** |
47
+
48
+ ## Data Splits
49
+
50
+ | Split | Images | Percentage |
51
+ |---|---|---|
52
+ | Train | 96,134 | ~80% |
53
+ | Validation | 12,333 | ~10% |
54
+ | Test | 11,874 | ~10% |
55
+
56
+ ## Folder Structure
57
+
58
+ Large folders are distributed as **zip archives** (HF enforces a 10,000-file-per-directory limit).
59
+
60
+ ```
61
+ train/
62
+ images.zip # RGB images (720 x 960) as .jpg
63
+ masks.zip # Segmentation masks (single-channel, class IDs 0-16) as .png
64
+ mmseg_masks.zip # Segmentation masks formatted for MMSegmentation as .png
65
+ xml.zip # VOC-format bounding box annotations as .xml
66
+ train_data.csv # Metadata: img, species, height, week
67
+ val/
68
+ images.zip
69
+ masks.zip
70
+ mmseg_masks.zip
71
+ xml.zip
72
+ val_data.csv
73
+ test/
74
+ images.zip
75
+ masks.zip
76
+ mmseg_masks.zip
77
+ xml.zip
78
+ test_data.csv
79
+ ```
80
+
81
+ ### After Extracting
82
+
83
+ Each zip extracts into a flat list of files. For example, `train/images.zip` contains:
84
+ ```
85
+ ABUTH_week_10_IMG_1656_frame_0.jpg
86
+ ABUTH_week_10_IMG_1656_frame_1.jpg
87
+ SETFA_week_8_IMG_1344_frame_13.jpg
88
+ ...
89
+ ```
90
+
91
+ ### File Naming Convention
92
+
93
+ All files follow the pattern: `{SPECIES}_week_{WEEK}_IMG_{VIDEO_ID}_frame_{FRAME}.{ext}`
94
+
95
+ Example: `SETFA_week_8_IMG_1344_frame_13.jpg`
96
+
97
+ ### CSV Metadata
98
+
99
+ Each split CSV contains columns:
100
+
101
+ | Column | Type | Description |
102
+ |---|---|---|
103
+ | `img` | string | Image filename (e.g., `ABUTH_week_10_IMG_1656_frame_0.jpg`) |
104
+ | `species` | string | Weed species EPPO code (e.g., `ABUTH`, `SETFA`) |
105
+ | `height` | float | Plant height in centimeters (0.2 - 155.0 cm) |
106
+ | `week` | int | Growth stage week (1 - 11) |
107
+
108
+ ### Segmentation Mask Values
109
+
110
+ | Pixel Value | Class |
111
+ |---|---|
112
+ | 0 | Background |
113
+ | 1 | ABUTH (Velvetleaf) |
114
+ | 2 | AMAPA (Palmer Amaranth) |
115
+ | 3 | AMARE (Redroot Pigweed) |
116
+ | 4 | AMATA / AMATU (Tall Waterhemp) |
117
+ | 5 | AMBEL (Common Ragweed) |
118
+ | 6 | CHEAL (Common Lambsquarters) |
119
+ | 7 | CYPES (Yellow Nutsedge) |
120
+ | 8 | DIGSA (Large Crabgrass) |
121
+ | 9 | ECHCG (Barnyardgrass) |
122
+ | 10 | ERICA (Horseweed) |
123
+ | 11 | PANDI (Fall Panicum) |
124
+ | 12 | SETFA (Giant Foxtail) |
125
+ | 13 | SETPU (Yellow Foxtail) |
126
+ | 14 | SIDSP (Prickly Sida) |
127
+ | 15 | SORHA (Johnsongrass) |
128
+ | 16 | SORVU (Shattercane) |
129
+
130
+ ## Weed Species Summary
131
+
132
+ | EPPO Code | Scientific Name | Max Height (cm) | Growth Rate (cm/week) | Category |
133
+ |---|---|---|---|---|
134
+ | AMATU | *Amaranthus tuberculatus* | 155.0 | 13.72 | Fast |
135
+ | SORHA | *Sorghum halepense* | 121.0 | 14.06 | Fast |
136
+ | SETFA | *Setaria faberi* | 124.0 | 11.75 | Fast |
137
+ | SORVU | *Sorghum bicolor* | 100.0 | 9.84 | Medium |
138
+ | PANDI | *Panicum dichotomiflorum* | 87.0 | 8.40 | Medium |
139
+ | SETPU | *Setaria pumila* | 99.0 | 8.20 | Medium |
140
+ | DIGSA | *Digitaria sanguinalis* | 77.0 | 7.53 | Medium |
141
+ | ECHCG | *Echinochloa crus-galli* | 80.0 | 7.38 | Medium |
142
+ | SIDSP | *Sida spinosa* | 69.0 | 6.77 | Medium |
143
+ | AMARE | *Amaranthus retroflexus* | 75.0 | 6.86 | Medium |
144
+ | ABUTH | *Abutilon theophrasti* | 72.0 | 6.32 | Medium |
145
+ | AMBEL | *Ambrosia artemisiifolia* | 71.0 | 6.19 | Medium |
146
+ | AMAPA | *Amaranthus palmeri* | 62.0 | 5.66 | Slow |
147
+ | CYPES | *Cyperus esculentus* | 56.0 | 5.42 | Slow |
148
+ | CHEAL | *Chenopodium album* | 30.0 | 2.86 | Slow |
149
+ | ERICA | *Erigeron canadensis* | 17.3 | 1.70 | Slow |
150
+
151
+ ## Data Collection
152
+
153
+ - **Location**: SIU Horticulture Research Center greenhouse, Southern Illinois University Carbondale, USA
154
+ - **Equipment**: iPhone 15 Pro Max positioned 1.5 feet above specimens
155
+ - **Capture**: 360-degree video at 1440 x 1920 resolution, 30 FPS
156
+ - **Environment**: 1000W HPS grow lights, 30-32 degree C
157
+ - **Preprocessing**: Temporal downsampling (every 2nd frame), spatial downscaling to 720 x 960
158
+ - **Annotation**: SAM2-Hiera-L semi-automatic segmentation with manual verification and correction
159
+ - **Height**: 325 manual weekly measurements (0.2 - 155 cm)
160
+
161
+ ## Benchmark Results (WeedSense Model)
162
+
163
+ | Task | Metric | Value |
164
+ |---|---|---|
165
+ | Segmentation | mIoU | 89.78% |
166
+ | Segmentation | mF1 | 94.54% |
167
+ | Height Estimation | MAE | 1.67 cm |
168
+ | Height Estimation | RMSE | 2.32 cm |
169
+ | Height Estimation | R squared | 0.9941 |
170
+ | Growth Stage | Accuracy | 99.99% |
171
+ | Growth Stage | F1 | 99.99% |
172
+
173
+ ## Usage
174
+
175
+ ```python
176
+ from huggingface_hub import snapshot_download, hf_hub_download
177
+ import zipfile, os
178
+
179
+ # Download entire dataset
180
+ local_dir = snapshot_download(repo_id="baselab/weedsense", repo_type="dataset")
181
+
182
+ # Extract a zip file
183
+ with zipfile.ZipFile(os.path.join(local_dir, "train", "images.zip"), "r") as z:
184
+ z.extractall(os.path.join(local_dir, "train", "images"))
185
+
186
+ # Load metadata
187
+ import pandas as pd
188
+ train_csv = hf_hub_download(
189
+ repo_id="baselab/weedsense",
190
+ repo_type="dataset",
191
+ filename="train/train_data.csv",
192
+ )
193
+ df = pd.read_csv(train_csv)
194
+ print(df.head())
195
+ # img species height week
196
+ # 0 ABUTH_week_10_IMG_1656_frame_0.jpg ABUTH 50.0 10
197
+ # 1 ABUTH_week_10_IMG_1656_frame_1.jpg ABUTH 50.0 10
198
+ ```
199
+
200
+ ## Citation
201
+
202
+ If you use this dataset, please cite:
203
+
204
+ ```bibtex
205
+ @inproceedings{sarker2025weedsense,
206
+ title={Weedsense: Multi-task learning for weed segmentation, height estimation, and growth stage classification},
207
+ author={Sarker, Toqi Tahamid and Ahmed, Khaled R and Islam, Taminul and Rankrape, Cristiana Bernardi and Gage, Karla},
208
+ booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
209
+ pages={7180--7190},
210
+ year={2025}
211
+ }
212
+ ```
213
+
214
+ ## License
215
+
216
+ This dataset is released under the [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license. Commercial use is not permitted. Derivative works must use the same license.