Table 4 The detail of three different image size of 256*256, 512*512, 1024*1024 datasets for training, validation and testing generated by our image synthetic method
From: High-throughput soybean seeds phenotyping with convolutional neural networks and transfer learning
Image size | Seed count | Large dataset | Small dataset | Test dataset | Generation time/min |
|---|---|---|---|---|---|
Train./Val. | Train./Val. | ||||
256*256 | 50–80 | 1000/100 | 200/20 | 200 | 274 |
512*512 | 80–100 | 1000/100 | 200/20 | 200 | 487 |
1024*1024 | 100–120 | 1000/100 | 200/20 | 200 | 575 |