Pour mesurer la manière dont le sys

Pour mesurer la manière dont le système ordonne des images pertinentes dans le résultat retourné à l’utilisateur, j’utilise la mesure numérique très populaire dans la communauté de RIC. C’est la précision moyenne (Average Precision).

Notons cependant qu’avec un scope k donné, la précision avec les k premières images retournées (dénotée par P@k) est proportionnelle au rappel (R@k) au même scope.

La précision moyenne pour une requête est calculée comme l’aire sous la courbe de précision-rappel en moyennant les précisions à chaque image pertinente retournée. La moyenne arithmétique de la précision moyenne calculée sur un nombre de différentes requêtes est appelée le MAP (Mean Average Precision).

Je calcule la précision aux 10, 20, 50, 100 et 200 premières images retournées pour les bases Caltech-4 et Caltech-101. Dans la case de la base Unbench, je calcule la précision aux 3 premières images puisqu’il n’y a que 3 images pertinentes pour une requête.

Kết quả (Anh) 1:

To measure the way in which the system directs images relevant in the output returned to the user, I use the popular digital measurement in the community of RIC. This is the average accuracy (Average Precision).Note, however, that with a given k, with k precision scope first returned images (denoted by P@k) is proportional to the callback (R@k) in the same scope.The average accuracy for a query is calculated as the area under the precision-recall curve by averaging each reverse relevant image information. The average arithmetic mean of the average accuracy calculated on a number of different queries is called MAP (Mean Average Precision).I calculate the accuracy at the 10, 20, 50, 100 and 200 first images returned for the Caltech-4 bases and Caltech-101. In the box of the Unbench base, I calculate the accuracy in the first 3 images since there are only 3 relevant images for a query.

Kết quả (Anh) 2:

To measure how the system directs relevant images in the results returned to the user, I use the very popular digital measurement in the RIC community. This is the average precision (Average Precision). Note, however, that with a given k scope, accuracy with the first images returned k (denoted by P @ k) is proportional to the recall (R @ k) at the same scope. The average accuracy for a query is calculated as the area under the precision-recall curve by averaging the relevant information to each mirror image. The arithmetic mean of the average precision computed on a number of different queries is called the MAP (Mean Average Precision). I calculate accuracy at 10, 20, 50, 100 and 200 first images returned to Caltech-4 bases and Caltech- 101. In the case of the base Unbench, I calculate accuracy at the first 3 images since there are only three relevant images for a query.

Kết quả (Anh) 3:

To measure the way in which the orderly system of relevant images in the result returned to the user, I used the digital measurement very popular in the community of RIC. This is the average accuracy (Average Precision) .

Note however that with a scope k given,The accuracy with the k first images returned (denoted by P@k) is proportional to the reminder (R@k) at the same scope.

The average accuracy for a query is calculated as the area under the curve of precision-recall in with the clarifications to each relevant image returned.The arithmetic average of the average accuracy calculated on a number of different queries is called the MAP (Mean Average Precision) .

I calculated the accuracy to 10, 20, 50, 100, and 200 first images returned for the databases Caltech-4 and Caltech-101. In the box of the database Unbench,I calculated the accuracy to 3 first images since there are only 3 images relevant to a query.