SegFormer_b2_10

This model is a fine-tuned version of nvidia/segformer-b2-finetuned-cityscapes-1024-1024 on an unknown dataset. It achieves the following results on the evaluation set:

• eval_loss: 0.8097
• eval_mean_iou: 0.7782
• eval_mean_accuracy: 0.8662
• eval_overall_accuracy: 0.9610
• eval_accuracy_road: 0.9903
• eval_accuracy_sidewalk: 0.9398
• eval_accuracy_building: 0.9628
• eval_accuracy_wall: 0.6563
• eval_accuracy_fence: 0.6967
• eval_accuracy_pole: 0.7388
• eval_accuracy_traffic light: 0.8695
• eval_accuracy_traffic sign: 0.8842
• eval_accuracy_vegetation: 0.9628
• eval_accuracy_terrain: 0.7418
• eval_accuracy_sky: 0.9803
• eval_accuracy_person: 0.9096
• eval_accuracy_rider: 0.7585
• eval_accuracy_car: 0.9782
• eval_accuracy_truck: 0.8827
• eval_accuracy_bus: 0.9472
• eval_accuracy_train: 0.8521
• eval_accuracy_motorcycle: 0.8134
• eval_accuracy_bicycle: 0.8929
• eval_iou_road: 0.9845
• eval_iou_sidewalk: 0.8702
• eval_iou_building: 0.9233
• eval_iou_wall: 0.5943
• eval_iou_fence: 0.5845
• eval_iou_pole: 0.5944
• eval_iou_traffic light: 0.6817
• eval_iou_traffic sign: 0.7808
• eval_iou_vegetation: 0.9255
• eval_iou_terrain: 0.6590
• eval_iou_sky: 0.9506
• eval_iou_person: 0.7943
• eval_iou_rider: 0.5888
• eval_iou_car: 0.9484
• eval_iou_truck: 0.8282
• eval_iou_bus: 0.8676
• eval_iou_train: 0.8049
• eval_iou_motorcycle: 0.6449
• eval_iou_bicycle: 0.7604
• eval_runtime: 188.4044
• eval_samples_per_second: 2.654
• eval_steps_per_second: 0.663
• epoch: 34.4086
• step: 6400

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 100
• mixed_precision_training: Native AMP

Framework versions

• Transformers 4.48.1
• Pytorch 2.1.2+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0