Add aircraft Swin-T serial 107 (vit_ce_448)

README.md

@@ -12,7 +12,7 @@ tags:
 
 These are the checkpoints behind *Fine-Grained Image Recognition from Scratch with Teacher-Guided Data Augmentation* ([arXiv:2507.12157](https://arxiv.org/abs/2507.12157)). Students trained from scratch learn from a fine-grained teacher (a ResNet-101 trained with counterfactual attention learning, CAL), which also guides data-aware augmentation of the student's inputs. Code: [arkel23/TGDA](https://github.com/arkel23/TGDA).
 
-249 checkpoints, one per configuration, each the last epoch of one training seed. Each file
+250 checkpoints, one per configuration, each the last epoch of one training seed. Each file
 is a `torch.save` dict with `config` (the full training configuration), `model` (the state dict),
 `accuracy` and `epoch`, with no optimizer state. File names are the runs' experiment-log names,
 ending in the serial. Load them with [fgir-zoo](https://github.com/arkel23/fgir-zoo). The
@@ -28,7 +28,7 @@ ending in the serial. Load them with [fgir-zoo](https://github.com/arkel23/fgir-
 | `vitfs_224_t448` | 102 | Teacher at 448 px, student at 224 px | 15 |
 | `vitfs_224` | 103 | Teacher and student at 224 px | 9 |
 | `vitfs_128` | 109 | ViTFS tiny/micro/nano students at 128 px from a 448 px teacher | 8 |
-| `vit_ce_448` | 107 | The serial-101 models trained with cross-entropy only (no teacher) | 35 |
+| `vit_ce_448` | 107 | The serial-101 models trained with cross-entropy only (no teacher) | 36 |
 | `vitfs_transfer` | 104-106 | A TGDA-trained student transferred to two new datasets with cross-entropy | 8 |
 | `ablations_128` | 212, 214, 221-223, 241-248, 251, 261-263, 271-278, 281 | Ablations: TGDA without extra regularization, one- vs two-stage training, augmentation type, teacher type, transfer | 73 |
 | `scratch_vs_pt_448` | 320, 321, 331 | 200 epochs from scratch with cross-entropy or KD, and ImageNet-pretrained with KD | 9 |
@@ -150,7 +150,7 @@ Each value is one seed; the paper reports its own aggregates, so its tables can
 | `nest_tiny` | 82.99 | 81.23 | 71.13 |
 | `pvt_v2_b0` | 86.41 | 89.76 | 75.09 |
 | `pvt_v2_b1` | 85.63 |  |  |
-| `swin_tiny_patch4_window7_224` |  | 63.26 |  |
+| `swin_tiny_patch4_window7_224` | 65.80 | 63.26 |  |
 | `vit_t16` | 45.60 | 26.38 | 30.81 |
 | `vitfs_tiny_patch16_gap_224` |  |  | 0.78 |
 | `vitfs_tiny_patch16_gap_reg4_dinov2_bn_init` |  | 88.41 | 69.92 |

manifest.csv

@@ -128,6 +128,7 @@ vit_ce_448/aircraft_mobilevitv2_125_107,vit_ce_448,aircraft_mobilevitv2_125_107,
 vit_ce_448/aircraft_nest_tiny_107,vit_ce_448,aircraft_nest_tiny_107,aircraft_nest_tiny_107.pth,vit_ce_448/aircraft_nest_tiny_107.pth,ERISLab/TGDA,aircraft,nest_tiny,,,False,107,0,448,100,82.99,paper,True,76a4f6b0bcb59e28b02cc02a24d31affec03a7206f8b5dfed0bb19c6cb00898c,73241030
 vit_ce_448/aircraft_pvt_v2_b0_107,vit_ce_448,aircraft_pvt_v2_b0_107,aircraft_pvt_v2_b0_107.pth,vit_ce_448/aircraft_pvt_v2_b0_107.pth,ERISLab/TGDA,aircraft,pvt_v2_b0,,,False,107,0,448,100,86.41,paper,True,d30724dca04fb9b9f968ae92568b7eb6d09590e62454a64fe44803165d9df3fd,13814633
 vit_ce_448/aircraft_pvt_v2_b1_107,vit_ce_448,aircraft_pvt_v2_b1_107,aircraft_pvt_v2_b1_107.pth,vit_ce_448/aircraft_pvt_v2_b1_107.pth,ERISLab/TGDA,aircraft,pvt_v2_b1,,,False,107,0,448,100,85.63,paper,True,0392040ac03ee0b3ba3cfe09ab599994d40ad613cef85e07e50e8bd82ecfa501,54262057
+vit_ce_448/aircraft_swin_tiny_patch4_window7_224_107,vit_ce_448,aircraft_swin_tiny_patch4_window7_224_107,aircraft_swin_tiny_patch4_window7_224_107.pth,vit_ce_448/aircraft_swin_tiny_patch4_window7_224_107.pth,ERISLab/TGDA,aircraft,swin_tiny_patch4_window7_224,,,False,107,0,448,100,65.8,paper,True,f5261f69b3890890eac06605b79ed6a37e7642b774d0ccb04df3982f387e8b3d,110464090
 vit_ce_448/aircraft_vit_t16_107,vit_ce_448,aircraft_vit_t16_107,aircraft_vit_t16_107.pth,vit_ce_448/aircraft_vit_t16_107.pth,ERISLab/TGDA,aircraft,vit_t16,,,False,107,0,448,100,45.6,paper,True,dc3fe4fd26825ce0b11b7703093f6d22897240788773059fc3816d2758421f54,22704830
 vit_ce_448/aircraft_vitfs_tiny_patch16_gap_reg4_dinov2_init_107,vit_ce_448,aircraft_vitfs_tiny_patch16_gap_reg4_dinov2_init_107,aircraft_vitfs_tiny_patch16_gap_reg4_dinov2_init_107.pth,vit_ce_448/aircraft_vitfs_tiny_patch16_gap_reg4_dinov2_init_107.pth,ERISLab/TGDA,aircraft,vitfs_tiny_patch16_gap_reg4_dinov2_init,,,False,107,0,448,100,76.84,paper,True,7a9cfc7bb77f9732c59fc2ce4cb5bda0918447b2e8f5bd8736472875bb1fa94b,23174911
 vit_ce_448/cars_efficientvit_m2_107,vit_ce_448,cars_efficientvit_m2_107,cars_efficientvit_m2_107.pth,vit_ce_448/cars_efficientvit_m2_107.pth,ERISLab/TGDA,cars,efficientvit_m2,,,False,107,0,448,196,83.75,paper,True,bf04cec4c54ab9300d525fb97d03a25cc5947d32b0adaddd0101866d6b41fcdf,16410069

vit_ce_448/aircraft_swin_tiny_patch4_window7_224_107.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5261f69b3890890eac06605b79ed6a37e7642b774d0ccb04df3982f387e8b3d
+size 110464090