Hey, I'm new here. I'm a final year Electrical and Electronics university student and for my FYP I'm creating an object detection app that integrates YOLOv5 model into a flutter app. I have trained my YOLOv5 model and successfully converted it into .tflite, but when I integrate my model into the flutter app, there is no output. Does anyone know the reason why? Could not find much resources on this topic as well. This is my first time creating an app and I really need some help.
For further actions, you may consider blocking this person and/or reporting abuse
Top comments (1)
"Hey! Integrating YOLOv5 with Flutter via TFLite can be tricky. A few things to check:
Input Shape/Format: Does your Flutter app pre-process images (resize, normalize) exactly like during YOLOv5 training? Even small mismatches can silence the model.
Output Parsing: YOLOv5's TFLite output is usually a flattened tensor. You might need to reshape it and apply confidence/NMS thresholds. Add print(rawOutput) right after inference to debug.
Metadata: If using tflite_flutter, verify your .tflite has metadata (use Netron to inspect). No metadata = silent fails sometimes.
Pro Tip: Test your .tflite standalone in a Python script first (using tf.lite.Interpreter) to isolate the issue.
Got a GitHub repo? Sharing code snippets (especially the inference part) would help! Also, consider posting this on the Ultralytics YOLOv5 issues page – they’re super responsive.