DAGSM: Disentangled Avatar Generation with GS-enhanced Mesh

Table of contents:

1. Animation videos of our avatars

Avatar 1: "A chubby african woman wearing a green long sleeve long straight silk dress"

Avatar 2: "A beautiful caucasian woman with short hair wearing a brown coat with gray t-shirt inside, skinny blue jeans and white board shoes"

Avatar 3: "A chinese teenage girl wearing a crew neck blue denim sleeveless A-skirt, gray leggings and white board shoes"

Avatar 4: "A middle-aged white man wearing a gray cashmere crewneck sweater, blue jeans and black leather shoes"

2. Clothing replacement results

Garments: brown coat, gray t-shirt, skinny blue jeans, white board shoes, pink hoodie, blue shirt, white lace sleeveless dress, khaki tailored trousers.

3. Muanual texture editing

PS: Click to play/stop the videos, drag to compare results.

Original Avatar Vs. After Muanual texture editing

4. Comparisons with baselines

We compare our method with three baselines that adopt three typical 3D representations, including:

Avatar 1: "A beautiful caucasian woman with short hair wearing a brown coat with gray t-shirt inside, skinny blue jeans and white board shoes"

Avatar 2: "A chubby african woman wearing a green long sleeve long straight silk dress"

Avatar 3: "A chinese teenage girl wearing a crew neck blue denim sleeveless A-skirt, gray leggings and white board shoes"

Avatar 4: "A middle-aged white man wearing a gray cashmere crewneck sweater, blue jeans and black leather shoes"

Avatar 5: "A african man wearing a blue Hawaiian shirt and gold Lakers basketball shorts"

Avatar 6: "A beautiful caucasian woman with short hair wearing a pink hoodie, skinny blue jeans and white board shoess"

Avatar 7: "A beautiful caucasian woman with short hair wearing a blue shirt, khaki tailored trousers and white board shoes"

Avatar 8: "A beautiful caucasian woman with short hair wearing a white lace sleeveless dress and khaki tailored trousers inside"