UPGRADE YOUR BROWSER

We have detected your current browser version is not the latest one. Xilinx.com uses the latest web technologies to bring you the best online experience possible. Please upgrade to a Xilinx.com supported browser:Chrome, Firefox, Internet Explorer 11, Safari. Thank you!

cancel
Showing results for 
Search instead for 
Did you mean: 
669 Views
Registered: ‎01-17-2019

F0117 20:26:37.153761 15086 bbox_util.cu:824] Check failed: match_index[p] < gt_bboxes.size() (1 vs. 1)

When I use the decent.sh, I have this error in the figure.

Screenshot from 2019-01-17 20-27-02.png

I have modified the prototxt file and trained the model with PASCAL VOC dataset as: https://forums.xilinx.com/t5/Deephi-DNNDK/SSD-support-in-DNNC/m-p/896797. And I have upload related files in Attached.

I really need your help to solve this problem. Thank you very much!

0 Kudos
3 Replies
Xilinx Employee
Xilinx Employee
626 Views
Registered: ‎02-17-2011

Re: F0117 20:26:37.153761 15086 bbox_util.cu:824] Check failed: match_index[p] < gt_bboxes.size() (1 vs. 1)

Hi @howtocodewang_,

I believe you need to remove your MultiBoxLoss layer as this is likely causing the issues.  I would just remove it and then retry.  

It also looks like you have two AnnotatedData input layers (one for test and one for train).  For the quantization/compilation process, I would remove the "train" AnnotatedData layer and use something like the following instead to specify a path to the calibration data (change the paths to point to your calibration.txt and images):


layer {
name: "data"
type: "ImageData"
top: "data"
top: "label"
include {
phase:TRAIN
}
transform_param {
crop_size: 300
mean_value: 104.0
mean_value: 117.0
mean_value: 123.0
resize_param {
prob: 1.0
resize_mode: WARP
height: 300
width: 300
interp_mode: LINEAR
interp_mode: AREA
interp_mode: NEAREST
interp_mode: CUBIC
interp_mode: LANCZOS4
}
}
image_data_param {
source: "./data/imagenet_320/calibration.txt"
root_folder:"./data/imagenet_320/calibration_images/"
batch_size: 10
shuffle: true
}
}

0 Kudos
604 Views
Registered: ‎01-17-2019

Re: F0117 20:26:37.153761 15086 bbox_util.cu:824] Check failed: match_index[p] < gt_bboxes.size() (1 vs. 1)

Hi @jcory,

Your solution is really helpful for me. decent.sh can work after removing my MultiBoxLoss layer. Thanks a lot.

However, I have another question when I try to deploy the quantized model (the .elf file) on the ZCU102 board. I modified the samples called video_analysis. But I got this error.

IMG_8883 2.JPG

I provide the modified main.cc file. Can you give me more information and figure out my faults.

Thanks!

0 Kudos
601 Views
Registered: ‎01-17-2019

Re: F0117 20:26:37.153761 15086 bbox_util.cu:824] Check failed: match_index[p] < gt_bboxes.size() (1 vs. 1)

@jcory This attached is the modified main.cc file.
0 Kudos