Deployment failed for GroveVIsion

At Deployment and firmware build failed …
/opt/arc_gnu/arc_gnu_ei_prebuilt_minimal_elf32_le_linux_install/bin/…/lib/gcc/arc-elf32/10.2.0/…/…/…/…/arc-elf32/bin/ld: obj_socket_24/gnu_arcem9d_wei_r16/WEI_FW_gnu_arcem9d_wei_r16.elf section .read_only_data' will not fit in region SYSTEM0’
/opt/arc_gnu/arc_gnu_ei_prebuilt_minimal_elf32_le_linux_install/bin/…/lib/gcc/arc-elf32/10.2.0/…/…/…/…/arc-elf32/bin/ld: region `SYSTEM0’ overflowed by 942508 bytes
Memory region Used Size Region Size %age Used
ICCM0: 0 GB 64 KB 0.00%
ICCM1: 227136 B 320 KB 69.32%
BOOT: 0 GB 130 KB 0.00%
SYSTEM0: 2382252 B 1406 KB 165.46%
DCCM: 20768 B 248 KB 8.18%
.stack: 0 GB 8 KB 0.00%
XCCM: 32 KB 32 KB 100.00%
YCCM: 32 KB 32 KB 100.00%
collect2: error: ld returned 1 exit status
make: *** [options/ obj_socket_24/gnu_arcem9d_wei_r16/WEI_FW_gnu_arcem9d_wei_r16.elf] Error 1
Application exited with code 2

Creating deployment failed

Job failed (see above)

Project ID:

Context/Use case:
Just testing workflow to understand the platform, evaluating for industrial dev in future

Looks like the program is bigger than the flash avaliable. I checked the website but could not find the spec for memory. Based on the error text it appears the device has 1.4MB and you program is 2.4MB.

Assuming this is a vision model, try reducing the Impulse design Imgae width and height to something smaller like 96x96.

Hi mate. Thanks. I chose the “smallest” model MobileNetv1 96x96 and managed to deploy and copy the firmware.uf2 over to my board. Then I ran

^C(base) stanley@stanLG:~/BouffaloLabDev$ edge-impulse-run-impulse --debug
Edge Impulse impulse runner v1.18.1
[SER] Connecting to /dev/ttyACM0
[SER] Serial is connected, trying to read config…
[SER] Retrieved configuration
[SER] Device is running AT command version 1.7.0

Want to see a feed of the camera and live classification in your browser? Go to

[SER] Started inferencing, press CTRL+C to stop…
Predictions (DSP: 2 ms., Classification: 87 ms., Anomaly: 0 ms.):

but the model downloaded seems to be not what I trained … it was classifying most of the image features as face(1.0)

oK. I reflashed again with the firmware.uf2, this time, no more face but not able to classify either paper or barcode. In the Live Classification, the model was at least able to ID …

What’s up? Sorry … steep learning curve here …

It looks like the Impulse is confused. By that I mean on the left side there is:

  • barcode,
  • paper, and
  • uncertain.

And on the right side:

  • barcode
  • paper
  • classified ← What is this?
  • classification 0 ← What is this?

Look thru all your data and check the Labels to make sure you only have barcode, paper, and uncertain. Then retrain and re-deploy.