# Model Evaluation Report ## Model: simpler_cnn - Test accuracy: 0.8286 ``` precision recall f1-score support No Overlay 0.87 0.90 0.89 154 Has Overlay 0.70 0.62 0.66 56 accuracy 0.83 210 macro avg 0.78 0.76 0.77 210 weighted avg 0.82 0.83 0.83 210 ``` ![Confusion Matrix for simpler_cnn](confusion_matrix_simpler_cnn.png) ### False Positive Images:

data/preprocessed/basspro.capitalone.com.png

data/preprocessed/forums.macrumors.com.png

data/preprocessed/bonusoffers.samsclub.com.png

data/preprocessed/www.petiteimpact.com.png

data/preprocessed/committees.parliament.uk.png

data/preprocessed/pcmanager.microsoft.com.png

data/preprocessed/www.premierneurology.org.png

data/preprocessed/bbbmarketplacetrust.org.png

data/preprocessed/www.wallstreetoasis.com.png

data/preprocessed/newsroom.bankofamerica.com.png

data/preprocessed/www.globalsmallbusinessblog.com.png

### False Negative Images:

data/preprocessed/www.balenciaga.com.png

data/preprocessed/www.theguardian.com.png

data/preprocessed/www.samsungsds.com.png

data/preprocessed/toronto.ubisoft.com.png

data/preprocessed/flights.virginatlantic.com.png

data/preprocessed/www.visiteauclaire.com.png

data/preprocessed/macyfoundation.org.png

data/preprocessed/www.maisonmargiela.com.png

data/preprocessed/stadiumcustomkicks.com.png

data/preprocessed/privacy.samsung.com.png

## Model: mobilenetv2 - Test accuracy: 0.7381 ``` precision recall f1-score support No Overlay 0.74 0.99 0.85 154 Has Overlay 0.67 0.04 0.07 56 accuracy 0.74 210 macro avg 0.70 0.51 0.46 210 weighted avg 0.72 0.74 0.64 210 ``` ![Confusion Matrix for mobilenetv2](confusion_matrix_mobilenetv2.png) ### False Positive Images:

data/preprocessed/www.harmonsgrocery.com.png

### False Negative Images:

data/preprocessed/www.visitrehoboth.com.png

data/preprocessed/www.foodnetwork.com.png

data/preprocessed/iowariverlanding.com.png

data/preprocessed/secretlosangeles.com.png

data/preprocessed/www.manoloblahnik.com.png

data/preprocessed/www.skinlabmaffi.com.png

data/preprocessed/www.bigchillislandhouse.com.png

data/preprocessed/www.thesunnysideupblog.com.png

data/preprocessed/www.tlhplasticsurgery.com.png

data/preprocessed/www.littletikes.com.png

data/preprocessed/www.theatlantic.com.png

data/preprocessed/www.saintlouisgalleria.com.png

data/preprocessed/www.radiotimes.com.png

data/preprocessed/www.riverchasegalleria.com.png

data/preprocessed/www.patagoniaburlington.com.png

data/preprocessed/apparel.onepeloton.com.png

data/preprocessed/us.merch.larian.com.png

## Model: resnet50 - Test accuracy: 0.9619 ``` precision recall f1-score support No Overlay 0.95 1.00 0.97 154 Has Overlay 1.00 0.86 0.92 56 accuracy 0.96 210 macro avg 0.98 0.93 0.95 210 weighted avg 0.96 0.96 0.96 210 ``` ![Confusion Matrix for resnet50](confusion_matrix_resnet50.png) ### False Positive Images: ### False Negative Images:

data/preprocessed/www.marketwatch.com.png