Transfer Learning model aids in identifying the sexes of wild monomorphic birds: Outperforms Geometric Morphometrics

Geometric Morphometrics (GMM) is well-known for shape analysis. We standardized and verified GMM methods to include images taken under field conditions and compared the results with Transfer Learning. Dataset (CMF) including code for PPS Dataset (CMF_20) including code for procrustes ANOVA Dataset (CMF_40) including code for procrustes ANOVA Dataset of Painted Stork images (TPS and Slider) and code for GMM Dataset for Transfer Learning models (OrientationNet, PSNet) MATLAB codes with sample images README: Transfer Learning model aids in identifying the sexes of wild monomorphic birds: Outperforms Geometric Morphometrics Description of the data and file structure a) To perform the Paired Procrustes Superimposition for detecting shape change due to orientations. Files and variables File: PPS.r Description: The R code to perform the Paired Procrustes Superimposition to get the Procrustes distance and rotation angle (theta) against each pair of orientations is available. File: six.TPS Description: We used six.TPS, which includes all 17 images of the CMF varying in orientation from 0° to 80°. We compared CMF (0°) with their counterparts oriented along different degrees (5° to 80°) using Paired Procrustes superimpositions (PPS). We plotted the Procrustes distance and rotation angle (theta) obtained between each pair against their orientations. b) To perform Procrustes ANOVA. File: procANOVA_set1.r Description: The R code to perform the Procrustes ANOVA for the first subset 0° to 20° is available. File: twentyangle.TPS Description: Using twentyangle.TPS, we quantified the orientation differences as systematic random errors in the first subset 0° to 20°. File: procANOVA_set2.r Description: The R code to perform the Procrustes ANOVA for the second subset 20° to 40° is available. File: fortyangle.TPS Description: Using fortyangle.TPS, we quantified the orientation differences as systematic random errors between 20° to 40°. c) To perform the shape analysis on Painted Stork’s mandible. File: ps_slider.r Description: The slider for semi landmarks of the Painted Stork’s mandible is available. File: twentyeightlmsSexes.TPS Description: The landmarks and semi landmarks of the sexes of Painted Stork’s mandible is available. File: bookpsadult51.csv Description: The variable mentioned below are available in bookpsadult51.csv for the analysis between sexes and location of adult Painted Stork. Variables S No: Stages: Sex: Location: Degree: File: ps_tps.r Description: The landmarks and semi landmarks of all growth stages of Painted Stork’s mandible is available. File: twentyeightlms147.TPS Description: The landmarks and semi landmarks of all growth stages of Painted Stork’s mandible is available. File: factors147.csv Description: The variable mentioned below are available in factors147.csv for the analysis among the growth stages of Painted Storks. Variables S No: Stages: Sex: Location: Baldness Degree: Bill D: bill Curve: Stag: loc: d) To run the Transfer Learning models using MATLAB framework. 1. Image Classification File: Orientation.m Description: The MATLAB codes to run the OrientationNet model is available. File: OrientationNet.mat Description: The OrientationNet model helps to identify the orientation (six classes) in degrees of a given image. File: image-299x299.jpg Description: Sample image to identify the orientation (six classes) in degrees. 2. Object Detection File: Object_Detection.m Description: The MATLAB codes to run the PSNet model is available. File: PSNet.mat Description: The PSNet model helps to identify the sexes (Male and Female) of Painted Stork in a given image. File: _73A4154.jpg Description: Sample image to identify sexes of Painted Storks. ######### Code/software a) The R codes to perform the Paired Procrustes Superimposition are available in File: PPS.r b) The R codes to perform the Procrustes ANOVA are available in File: procANOVA_set1.r c) The R codes to perform the shape analysis on the mandibles of Painted Stork are available in File: ps_slider.r d) The sample images (File: image-299x299.jpg), MATLAB codes (File: Orientation.m; ObjectDetection.m) & models (File: OrientationNet, PSNet) are available.