Age and Gender Prediction using Deep cnns and Transfer Learning

Download 299.02 Kb. Pdf ko'rish
bet	3/13
Sana	22.12.2023
Hajmi	299.02 Kb.
	#127116

1 2 3 4 5 6 7 8 9 ... 13

Bog'liq
xudo xoxlasa tushadi99%, 3-labarotoriya ishi Saralash usul va algoritmlarini tadqiq qilis, cmd buyruqlari, Incremental model nima, 1matematik, word sAM 1 savol, Документ Microsoft Word (4), Ma\'ruzalar (2), ЛАБОРАТОРНАЯ РАБОТА N1, Dasturlash 2, Ariza, Qalandarova Gulshoda, 1648631455, 1650692784, 1651669892 (2)

2
Related Work
Initial works of age and gender prediction involved techniques based on ratios of dif-
ferent measurements of facial features such as size of eye, nose, distance of chin from
forehead, distance between the ears, angle of inclination, angle between locations [8].
Such methods were known as anthropometric methods.
Early methods were based on manual extraction of features such as PCA, LBP, Ga-
bor, LDA, SFP. These extracted features were then fed to classical ML models such as
SVMs, decision trees, logistic regression. Hu et al. [9] used the method of ULBP, PCA
& SVM for age estimation. Guo et al. [10] proposed a locally adjusted robust regression
(LARR) algorithm, which combines SVM and SVR when estimating age by first using
SVR to estimate a global age range, and then using SVM to perform exact age estima-
tion. The obvious down-side of such methods was that not only getting anthropometric
measurements was difficult but the models were not able to generalize because people
of different age and gender could have the same anthropometric measurements.
Recently the use of CNN for age and gender prediction has been widely adopted as
CNNs are pretty robust and give outstanding results when tested on face images with
occlusion, tilt, altered brightness. Such results have been attributed to its good ability

3
to extract features. This happens by convolving over the given image to generate invar-
iant features which are passed onto the next layer in a sequential fashion. It is this con-
tinual passing of information from one layer to the next that leads to CNNs being so
robust and supple to occlusions, brightness changes etc.
The first application of CNNs was the Le-Net-5 [11]. However, the actual boom in
using CNNs for age and gender prediction started after D-CNN [12] was introduced for
image classification tasks. Rothe et al. [13] proposed DEX: Deep EXpectation of Ap-
parent Age for age classification using an ensemble of 20 networks on the cropped faces
of IMDB-Wiki dataset. Another popular work includes combining features from Deep
CNN to features obtained from PCA done by Wang et al. [14].

Download 299.02 Kb.

1 2 3 4 5 6 7 8 9 ... 13

Download 299.02 Kb.

Pdf ko'rish