Skip to content

uisikdag/plantseedling

BranchesTags

Repository files navigation

Weed Classification Model Benchmark Report with Stratified 5-Fold Cross-Validation

Dataset Information

  • Total samples: 3000
  • Training+Validation samples: 2700
  • Test samples: 300
  • Classes: Black-grass, Charlock, Cleavers, Common Chickweed, Common wheat, Fat Hen, Loose Silky-bent, Maize, Scentless Mayweed, Shepherds Purse, Small-flowered Cranesbill, Sugar beet

Training Results

  • Cross-validation: Stratified 5-fold

  • Accuracy,Precision,Recall,F1-Score are calculated based on Test Set

  • Best Validation Accuracy is calculated as the accuracy of the fold with best value

  • Average Per Class Performance for each fold is calculated as the mean of Test Set results for all folds

  • Std Dev refers to the standard deviation of the Test Accuracy across different folds in the cross-validation process.

Model Performance Comparison

Model Accuracy Precision Recall F1-Score Best Val Acc Parameters (M) Std Dev Completed Folds
swinv2 97.4 97.56 97.4 97.4 97.3 86.91 0.65 5
convnext_large 97.33 97.45 97.33 97.33 97.74 196.25 0.6 5
nasnet_large 97.07 97.32 97.07 97.06 97.37 84.77 0.39 5
mobilenet_v3 96.93 97.13 96.93 96.93 97.04 4.22 0.25 5
davit 96.87 97.01 96.87 96.87 97.78 86.94 1.09 5
efficientnet_v3 96.73 96.83 96.73 96.73 97.11 20.19 0.71 5
inception_resnet_v2 96.67 96.83 96.67 96.68 97.59 54.32 0.6 5
resnet152 96.53 96.72 96.53 96.52 96.96 58.17 0.4 5
vgg19 95.47 95.6 95.47 95.47 95.44 139.63 0.81 5
coatnet_3 90.13 90.65 90.13 90.11 89.67 163.65 3.67 5
vit_large_patch16 58.33 59.71 58.33 56.92 61.56 303.31 17.5 5

Individual Model Results

vgg19

Architecture Details

  • Total Parameters: 139,630,412

Cross-Validation Results

Fold Best Val Acc Test Acc Precision Recall F1-Score
1 95.56% 95.33% 95.41% 95.33% 95.31%
2 95.19% 94.33% 94.39% 94.33% 94.32%
3 95.00% 96.00% 96.28% 96.00% 96.03%
4 95.19% 96.67% 96.78% 96.67% 96.67%
5 96.30% 95.00% 95.12% 95.00% 95.02%
Avg 95.44% 95.47% ± 0.81% 95.60% 95.47% 95.47%

Average Per-Class Performance

Class Precision Recall F1-Score
Black-grass 81.20% 86.40% 83.71%
Charlock 100.00% 100.00% 100.00%
Cleavers 100.00% 96.00% 97.96%
Common Chickweed 95.44% 100.00% 97.66%
Common wheat 92.05% 97.60% 94.64%
Fat Hen 96.94% 94.40% 95.59%
Loose Silky-bent 87.82% 80.80% 84.15%
Maize 99.23% 100.00% 99.61%
Scentless Mayweed 99.20% 97.60% 98.38%
Shepherds Purse 97.60% 96.80% 97.18%
Small-flowered Cranesbill 97.69% 100.00% 98.82%
Sugar beet 100.00% 96.00% 97.94%

resnet152

Architecture Details

  • Total Parameters: 58,168,396

Cross-Validation Results

Fold Best Val Acc Test Acc Precision Recall F1-Score
1 96.67% 97.00% 97.43% 97.00% 96.97%
2 96.85% 96.33% 96.57% 96.33% 96.32%
3 97.04% 97.00% 97.08% 97.00% 97.00%
4 97.41% 96.33% 96.44% 96.33% 96.33%
5 96.85% 96.00% 96.07% 96.00% 96.00%
Avg 96.96% 96.53% ± 0.40% 96.72% 96.53% 96.52%

Average Per-Class Performance

Class Precision Recall F1-Score
Black-grass 84.13% 91.20% 87.30%
Charlock 100.00% 100.00% 100.00%
Cleavers 100.00% 96.00% 97.96%
Common Chickweed 94.73% 100.00% 97.29%
Common wheat 95.41% 99.20% 97.25%
Fat Hen 100.00% 94.40% 97.11%
Loose Silky-bent 91.01% 82.40% 86.18%
Maize 99.23% 100.00% 99.61%
Scentless Mayweed 99.20% 96.80% 97.98%
Shepherds Purse 96.89% 99.20% 98.02%
Small-flowered Cranesbill 100.00% 100.00% 100.00%
Sugar beet 100.00% 99.20% 99.59%

inception_resnet_v2

Architecture Details

  • Total Parameters: 54,324,908

Cross-Validation Results

Fold Best Val Acc Test Acc Precision Recall F1-Score
1 98.15% 97.00% 97.15% 97.00% 97.00%
2 97.78% 96.33% 96.67% 96.33% 96.38%
3 96.48% 97.00% 97.14% 97.00% 97.01%
4 97.78% 97.33% 97.44% 97.33% 97.33%
5 97.78% 95.67% 95.75% 95.67% 95.66%
Avg 97.59% 96.67% ± 0.60% 96.83% 96.67% 96.68%

Average Per-Class Performance

Class Precision Recall F1-Score
Black-grass 84.99% 94.40% 89.41%
Charlock 100.00% 100.00% 100.00%
Cleavers 100.00% 96.00% 97.96%
Common Chickweed 93.30% 100.00% 96.53%
Common wheat 93.81% 96.00% 94.87%
Fat Hen 99.17% 92.80% 95.87%
Loose Silky-bent 93.00% 84.80% 88.71%
Maize 97.69% 100.00% 98.82%
Scentless Mayweed 100.00% 100.00% 100.00%
Shepherds Purse 100.00% 100.00% 100.00%
Small-flowered Cranesbill 100.00% 100.00% 100.00%
Sugar beet 100.00% 96.00% 97.94%

mobilenet_v3

Architecture Details

  • Total Parameters: 4,217,404

Cross-Validation Results

Fold Best Val Acc Test Acc Precision Recall F1-Score
1 97.78% 96.67% 97.22% 96.67% 96.62%
2 96.48% 96.67% 96.81% 96.67% 96.67%
3 96.67% 97.00% 97.09% 97.00% 97.01%
4 97.41% 97.33% 97.45% 97.33% 97.33%
5 96.85% 97.00% 97.09% 97.00% 97.01%
Avg 97.04% 96.93% ± 0.25% 97.13% 96.93% 96.93%

Average Per-Class Performance

Class Precision Recall F1-Score
Black-grass 85.82% 94.40% 89.78%
Charlock 100.00% 100.00% 100.00%
Cleavers 100.00% 96.00% 97.96%
Common Chickweed 94.02% 100.00% 96.91%
Common wheat 96.98% 99.20% 98.04%
Fat Hen 100.00% 94.40% 97.11%
Loose Silky-bent 94.06% 84.00% 88.50%
Maize 100.00% 100.00% 100.00%
Scentless Mayweed 99.23% 97.60% 98.38%
Shepherds Purse 97.69% 99.20% 98.42%
Small-flowered Cranesbill 100.00% 100.00% 100.00%
Sugar beet 97.75% 98.40% 98.02%

tf_efficientnetv2_s

Architecture Details

  • Total Parameters: 20,192,860

Cross-Validation Results

Fold Best Val Acc Test Acc Precision Recall F1-Score
1 97.04% 97.33% 97.41% 97.33% 97.34%
2 97.59% 96.33% 96.47% 96.33% 96.32%
3 96.67% 95.67% 95.73% 95.67% 95.66%
4 96.85% 97.67% 97.76% 97.67% 97.65%
5 97.41% 96.67% 96.78% 96.67% 96.66%
Avg 97.11% 96.73% ± 0.71% 96.83% 96.73% 96.73%

Average Per-Class Performance

Class Precision Recall F1-Score
Black-grass 86.87% 88.80% 87.72%
Charlock 100.00% 100.00% 100.00%
Cleavers 100.00% 96.00% 97.96%
Common Chickweed 96.15% 100.00% 98.04%
Common wheat 92.64% 100.00% 96.17%
Fat Hen 100.00% 95.20% 97.53%
Loose Silky-bent 89.38% 85.60% 87.31%
Maize 99.23% 100.00% 99.61%
Scentless Mayweed 99.23% 98.40% 98.79%
Shepherds Purse 98.46% 99.20% 98.81%
Small-flowered Cranesbill 100.00% 100.00% 100.00%
Sugar beet 100.00% 97.60% 98.78%

nasnetalarge.tf_in1k

Architecture Details

  • Total Parameters: 84,768,546

Cross-Validation Results

Fold Best Val Acc Test Acc Precision Recall F1-Score
1 98.15% 96.67% 96.83% 96.67% 96.66%
2 97.78% 97.67% 97.88% 97.67% 97.66%
3 96.67% 97.33% 97.64% 97.33% 97.32%
4 97.41% 96.67% 96.91% 96.67% 96.66%
5 96.85% 97.00% 97.36% 97.00% 97.00%
Avg 97.37% 97.07% ± 0.39% 97.32% 97.07% 97.06%

Average Per-Class Performance

Class Precision Recall F1-Score
Black-grass 84.27% 98.40% 90.77%
Charlock 100.00% 100.00% 100.00%
Cleavers 100.00% 95.20% 97.53%
Common Chickweed 96.92% 100.00% 98.43%
Common wheat 93.16% 96.80% 94.92%
Fat Hen 100.00% 96.00% 97.96%
Loose Silky-bent 98.14% 81.60% 89.08%
Maize 96.92% 100.00% 98.43%
Scentless Mayweed 99.23% 99.20% 99.20%
Shepherds Purse 99.23% 100.00% 99.61%
Small-flowered Cranesbill 100.00% 100.00% 100.00%
Sugar beet 100.00% 97.60% 98.78%

vit_large_patch16_224

Architecture Details

  • Total Parameters: 303,313,932

Cross-Validation Results

Fold Best Val Acc Test Acc Precision Recall F1-Score
1 21.67% 25.33% 29.37% 25.33% 22.38%
2 70.74% 63.67% 63.59% 63.67% 62.64%
3 64.07% 61.00% 62.49% 61.00% 59.78%
4 81.11% 77.67% 78.23% 77.67% 77.60%
5 70.19% 64.00% 64.86% 64.00% 62.17%
Avg 61.56% 58.33% ± 17.50% 59.71% 58.33% 56.92%

Average Per-Class Performance

Class Precision Recall F1-Score
Black-grass 61.48% 55.20% 57.08%
Charlock 62.25% 79.20% 69.24%
Cleavers 63.19% 74.40% 68.25%
Common Chickweed 56.97% 59.20% 52.02%
Common wheat 70.60% 75.20% 72.60%
Fat Hen 57.32% 44.80% 49.44%
Loose Silky-bent 67.13% 54.40% 57.49%
Maize 67.29% 64.80% 65.78%
Scentless Mayweed 53.96% 55.20% 51.51%
Shepherds Purse 35.10% 32.00% 32.83%
Small-flowered Cranesbill 57.28% 55.20% 54.09%
Sugar beet 63.93% 50.40% 52.66%

davit_base.msft_in1k

Architecture Details

  • Total Parameters: 86,941,708

Cross-Validation Results

Fold Best Val Acc Test Acc Precision Recall F1-Score
1 97.96% 98.33% 98.45% 98.33% 98.34%
2 97.78% 96.67% 96.80% 96.67% 96.68%
3 97.22% 97.33% 97.45% 97.33% 97.33%
4 98.52% 97.00% 97.10% 97.00% 97.00%
5 97.41% 95.00% 95.25% 95.00% 94.98%
Avg 97.78% 96.87% ± 1.09% 97.01% 96.87% 96.87%

Average Per-Class Performance

Class Precision Recall F1-Score
Black-grass 88.34% 91.20% 89.59%
Charlock 100.00% 100.00% 100.00%
Cleavers 100.00% 96.00% 97.96%
Common Chickweed 94.02% 100.00% 96.91%
Common wheat 91.39% 98.40% 94.68%
Fat Hen 100.00% 92.80% 96.24%
Loose Silky-bent 93.48% 88.80% 91.03%
Maize 99.23% 100.00% 99.61%
Scentless Mayweed 100.00% 97.60% 98.78%
Shepherds Purse 97.69% 100.00% 98.82%
Small-flowered Cranesbill 100.00% 100.00% 100.00%
Sugar beet 100.00% 97.60% 98.78%

swinv2_base_window8_256.ms_in1k

Architecture Details

  • Total Parameters: 86,906,116

Cross-Validation Results

Fold Best Val Acc Test Acc Precision Recall F1-Score
1 98.15% 97.67% 97.91% 97.67% 97.66%
2 97.96% 98.33% 98.47% 98.33% 98.33%
3 96.67% 97.67% 97.77% 97.67% 97.66%
4 96.67% 96.67% 96.79% 96.67% 96.67%
5 97.04% 96.67% 96.86% 96.67% 96.67%
Avg 97.30% 97.40% ± 0.65% 97.56% 97.40% 97.40%

Average Per-Class Performance

Class Precision Recall F1-Score
Black-grass 88.35% 96.00% 91.92%
Charlock 100.00% 100.00% 100.00%
Cleavers 100.00% 96.00% 97.96%
Common Chickweed 94.73% 100.00% 97.29%
Common wheat 93.28% 99.20% 96.14%
Fat Hen 100.00% 93.60% 96.68%
Loose Silky-bent 95.91% 87.20% 91.23%
Maize 99.23% 100.00% 99.61%
Scentless Mayweed 100.00% 99.20% 99.59%
Shepherds Purse 99.23% 100.00% 99.61%
Small-flowered Cranesbill 100.00% 100.00% 100.00%
Sugar beet 100.00% 97.60% 98.78%

convnext_large

Architecture Details

  • Total Parameters: 196,248,780

Cross-Validation Results

Fold Best Val Acc Test Acc Precision Recall F1-Score
1 97.78% 98.33% 98.47% 98.33% 98.33%
2 97.59% 97.00% 97.14% 97.00% 96.99%
3 97.59% 97.67% 97.69% 97.67% 97.67%
4 97.78% 96.67% 96.82% 96.67% 96.66%
5 97.96% 97.00% 97.16% 97.00% 97.00%
Avg 97.74% 97.33% ± 0.60% 97.45% 97.33% 97.33%

Average Per-Class Performance

Class Precision Recall F1-Score
Black-grass 86.26% 95.20% 90.49%
Charlock 100.00% 100.00% 100.00%
Cleavers 100.00% 96.00% 97.96%
Common Chickweed 97.69% 100.00% 98.82%
Common wheat 96.18% 98.40% 97.24%
Fat Hen 98.40% 97.60% 97.99%
Loose Silky-bent 94.70% 84.80% 89.45%
Maize 97.75% 100.00% 98.84%
Scentless Mayweed 100.00% 98.40% 99.18%
Shepherds Purse 98.46% 100.00% 99.22%
Small-flowered Cranesbill 100.00% 100.00% 100.00%
Sugar beet 100.00% 97.60% 98.78%

coatnet_3_rw_224.sw_in12k

Architecture Details

  • Total Parameters: 163,654,872

Cross-Validation Results

Fold Best Val Acc Test Acc Precision Recall F1-Score
1 94.26% 93.00% 93.41% 93.00% 93.07%
2 82.41% 84.00% 84.67% 84.00% 84.04%
3 92.78% 94.33% 95.09% 94.33% 94.25%
4 91.67% 91.00% 91.21% 91.00% 90.99%
5 87.22% 88.33% 88.86% 88.33% 88.22%
Avg 89.67% 90.13% ± 3.67% 90.65% 90.13% 90.11%

Average Per-Class Performance

Class Precision Recall F1-Score
Black-grass 84.76% 72.80% 77.74%
Charlock 93.58% 90.40% 91.88%
Cleavers 96.17% 91.20% 93.51%
Common Chickweed 85.82% 98.40% 91.58%
Common wheat 92.77% 92.00% 92.27%
Fat Hen 92.65% 88.80% 90.64%
Loose Silky-bent 76.98% 87.20% 81.51%
Maize 97.69% 100.00% 98.82%
Scentless Mayweed 91.75% 92.80% 92.13%
Shepherds Purse 84.67% 88.00% 86.29%
Small-flowered Cranesbill 93.66% 90.40% 91.91%
Sugar beet 97.28% 89.60% 93.10%

About

No description, website, or topics provided.

Resources

Readme

License

GPL-3.0 license
Activity

Stars

0 stars

Watchers

1 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

No packages published