How to implement batch predict?

[fanweiya]

Hi Ayoola, many thanks for creating this library.How do I implement batch predict?How can this library process hundreds of images at once?

[ayoolaolafenwa]

@fanweiya you will create a folder and put the images inside the folder. Perform batch prediction with this modified code.

import pixellib
from pixellib.instance import instance_segmentation
import cv2
#import glob to use it to retrieve image names
import glob


seg = instance_segmentation()
seg.load_model("mask_rcnn_coco.h5")

#get image names with glob
imgnames = sorted(glob.glob("pics/*.jpg"))

#loop through each image in the folder
for imgname in imgnames:
    #process images
   #segment each image consecutively through the loop
    segvalues, output = seg.segmentImage(imgname)
    
    # save output images
   #get image name with the corresponding extension and save them 
  #image saved like sample_output.jpg
    name, ext = os.path.splitext(imgname)
    imgname2 = imgname+"_output"+ext
    cv2.imwrite(imgname2, output) 

Note You did not specify whether you want to batch predict in instance segmentation or semantic segmentation, but the code for iterating through the folder of images is still the same for both, except you will replace the code for instance segmentation for semantic segmentation for performing batch prediction in semantic segmentation.

[fanweiya]

@ayoolaolafenwa I want to predict more than one image at a time,something like Keras model.predict_on_batch.

@fanweiya you will create a folder and put the images inside the folder. Perform batch prediction with this modified code.

import pixellib
from pixellib.instance import instance_segmentation
import cv2
#import glob to use it to retrieve image names
import glob


seg = instance_segmentation()
seg.load_model("mask_rcnn_coco.h5")

#get image names with glob
imgnames = sorted(glob.glob("pics/*.jpg"))

#loop through each image in the folder
for imgname in imgnames:
    #process images
   #segment each image consecutively through the loop
    segvalues, output = seg.segmentImage(imgname)
    
    # save output images
   #get image name with the corresponding extension and save them 
  #image saved like sample_output.jpg
    name, ext = os.path.splitext(imgname)
    imgname2 = imgname+"_output"+ext
    cv2.imwrite(imgname2, output) 

Note You did not specify whether you want to batch predict in instance segmentation or semantic segmentation, but the code for iterating through the folder of images is still the same for both, except you will replace the code for instance segmentation for semantic segmentation for performing batch prediction in semantic segmentation.

[ayoolaolafenwa]

@fanweiya this is presently the way to predict multiple images with pixellib and in the nearest future I may introduce a batch prediction function.