Garment From Mannequin Images

Overview & Supported Outputs

Ghost mannequin garment images or images of mannequins wearing the garment in general can be tansformed into GARMENT_2D assets for virtual try-on applications such as Fashion Image Generation and Virtual Fashion Try-On & Sizing.
- File Extension: .garment
- Content Service Type: GARMENT_2D
For more information on Output Asset Type Specifications, click here.

Image Quality Specifications

Image Properties

Image Property	Recommended Specification	Remarks
File Type	JPG (JPEG) or PNG 8-Bit	PICTOFiT Studio only recognizes these file types as input images for the `GARMENT_2D` asset type.
File Size	Ideal size is 500 KB or less (limit is 20 MB)	Images with large file size tend to affect site performance.
Image Width	1500 to 2500 pixels wide (1000 pixels minimum)	Images smaller than the recommended baseline will appear blurry or pixelated. The garment should use as much space as possible, reducing the white space to a minimum.
Resolution	Up to 60 MP (megapixel)	Image resolution can be calculated by multiplying the dimensions of the image and dividing the product by 1 million
Color Mode	RGB	CMYK color mode is used only for print
Color Profile	sRGB	The images will not look good on devices if the color profile is not sRGB*

Input Image Requirements

`GHOST MANNEQUIN SHOT`

A ghost mannequin image of the garment is required. This image of the garment is what we will primarily use to process the GARMENT_2D asset. Ghost mannequin images can either be the front or the back view of the garment.
Note that front view garment image are intended to work with front view avatars and vice versa (back view garment images and avatars) for a realistic result.

FRONT VIEW

BACK VIEW

What is a Ghost Mannequin Image?

The ghost mannequin is also known as invisible mannequin, 3D mannequin or hollow man.
It is a post-production technique for apparel product photography.

Brief Guide to Preparing a Ghost Mannequin Image

Take a photograph of a garment on a mannequin.
[Optional: only if parts are hidden] Take one or multiple of the parts the garment that were covered by the mannequin (e.g. collar backside, cuffs of the sleeves, etc.)
In Post-processing, remove the mannequin of the first image and reconstruct the hidden parts using the images captured in step 2.

A detailed guide is available here.

`REFERENCE IMAGE`

Although providing reference images is optional, it is encouraged to submit them because it helps us ensure the accuracy of the garment’s overall fit.

FRONT VIEW

BACK VIEW

(Optional) Tech Packs for Size Recommendation & Visualisation

To enable size recommendation & visualisation, we require sizing tables which are usually part of the tech packs of a garment. Since these come in different forms, please send us an example, so we can make sure that we can work with them.

Garment Quality Specifications

Category A | Ideal Product Image + Reference Image

Product images classified as Category A are processed with minimal correction and guarantee an optimal result.
Garments in Category A have a GHOST MANNEQUIN SHOT and a REFERENCE IMAGE.
Moreover, a product image belongs to this category if the following specifications are fulfilled:

Specification	Input Sample	Output Sample
The garment is completely worn by a ghost mannequin representing an actual human body. No part of the garment is captured flat or deflated.
The pose of the ghost mannequin is straight and symmetrical from left to right (A pose).
All parts of the garment are visible. Parts may touch but they should not overlap.
The garment has a high contrast to the background and is captured in optimal lighting, showing a sharp and distinct outline/edge.
All backsides are visible (e.g. collar).
Accessories are not fitted to the avatar but are still processed (background removal) as part of the whole outfit.

Category B | Extra effort but guaranteed ideal result

Product images classified as Category B do not fulfill the specifications listed in Category A.
Garments considered as Category B have a GHOST MANNEQUIN IMAGE and a REFERENCE IMAGE.
Moreover, a product image belongs to this category if the following specifications are present:

Issues

Input Sample

Output Sample

Partial Overlapping Parts
Parts of the garment are covered in the product image

Reasons for this can be:

The mannequin arms are too close to the torso. Especially wide sleeves/or wide shapes e.g. bell shape may result in overlaps. (example 1
Bows, belts, or other decorative elements that stick out and as a result cover another part of the garment (see pink dress as example)
The mannequin arms are not attached so that the sleeves are hanging down in the back/front.

Thin Sleeves
The sleeves of the garment are too thin.

Reasons for this can be:

The absence of arms which should be attached to the ghost mannequin, making the sleeves deflated
The way the garment is captured, the torso is more prominent, hiding the parts of the upper arm.

Low Contrast
The garment has low contrast to the background, especially for garments with intricate details such as fur, fringes, lace, etc.

Category C | Extra effort and/or un-guaranteed ideal result

Product images classified as Category C do not fulfill the specifications mentioned in Categories A and B. Furthermore, they may have the following limitations:
- The garments need correction but still does not guarantee an optimal result.
- No REFERENCE IMAGE image is provided and the accuracy of the fit cannot be guaranteed.
Issues in this category include the following:

Issues	Input Sample	Output Sample
Frontal Model The garment is worn by a model. Parts of the garment are covered and/or the silhouette of the garment does not match the pose of the avatar used.
Backside Missing The garment is missing it’s backside.
Hanger Shot The garment is captured on a hanger.
Flat The garment is captured without a ghost mannequin.
Deflated Sleeves Parts of the garment are deflated. (e.g. using a ghost mannequin without arms)
Poor Lighting Garments captured in poor lighting conditions.
Transparent Overlaps Transparent garments that are captured without a ghost mannequin where front and back parts overlap.

Not Processable

The following product images cannot be processed:

Fully Covered Large parts of the garments are covered. Missing information is impossible to replicate.
Model Pose The models in the image depict the following: The model is posing in a way that make the garment not fully visible. The model is posing in such a way that the garment is distorted (hands on pocket, hands holding accessories, hair covering the garment). The model is wearing an upper garment that is tucked in.
Sideways The garment is not frontal in the image
Unclear Especially for garments with no reference pictures (model wearing the actual garment) The fit is too complicated to guess
Multiple Garments Multiple garments that overlap (example 1) as well as garments that are separated but disproportional and not positioned properly (example 2)	Example 1 (Overlapping) Example 2 (Disproportional)
Folded Product images with garments that are bent and/or folded