Zombie exercise not working as expected

[nksubramanian]

Hi , when i run the comparaison cell below is what I get on the Week2 Zombie detection exercise.
237
(4,)

False
False
False
and later a 0.0 score

could you please help? thanks.

[TMosh]

Which course are you attending?

You’ve posted in “AI Discussions”, but that’s for general topics - not a specific course.

[nksubramanian]

Oh sorry. course is Advanced computervision with Tensorflow. and the issue is in Week 2 Programming assignment. Zombie Detector.

[TMosh]

I’ll move your thread to that forum area. This will make it more likely to be found by a mentor for that course.

[gent.spah]

Which cell is that in particular, this assignment is highly dependent from top to bottom and any mistake in previous cells can cause an issue.

Did you see what the person in the previous post used to solve it? There are also many posts solved for this assignment in this specialisation page, I would advice to go through them.

Also in case you still cannot find the problem I would advice redoing the assignment from the beginning because sometimes a second run helps in catching issues.

[nksubramanian]

Hi, thanks for getting back to me. i have checked and double checked. checked the checkpoint directory usage etc, all seems fine to me. Hence reaching out your team for assistance.

1. The first issue i face is in section 6.1 and 6.2. Cell 40&43, Not sure if this needs to be fixed or ignored.

Expected output:

tensorflow.python.training.tracking.util.Checkpoint

But i get tensorflow.python.checkpoint.checkpoint.Checkpoint

2. Second issue is, In 6.2, cell 4,
   vars(tmp_model_checkpoint)

i get the output as

{'_root': None,
 '_self_setattr_tracking': True,
 '_self_unconditional_checkpoint_dependencies': [TrackableReference(name=model, ref=<tensorflow.python.checkpoint.checkpoint.Checkpoint object at 0x7a3f3f470490>)],
 '_self_unconditional_dependency_names': {'model': <tensorflow.python.checkpoint.checkpoint.Checkpoint at 0x7a3f3f470490>},
 '_self_unconditional_deferred_dependencies': {},
 '_self_update_uid': -1,
 '_self_name_based_restores': set(),
 '_self_saveable_object_factories': {},
 '_kwargs': {'model': <tensorflow.python.checkpoint.checkpoint.Checkpoint at 0x7a3f3f470490>},
 '_async_checkpointer_impl': None,
 '_checkpoint_options': None,
 '_save_counter': None,
 '_save_assign_op': None,
 'model': <tensorflow.python.checkpoint.checkpoint.Checkpoint at 0x7a3f3f470490>,
 '_saver': <tensorflow.python.checkpoint.checkpoint.TrackableSaver at 0x7a3f3f470850>,
 '_attached_dependencies': None}

Among the variables of this checkpoint, you should see:

'_box_predictor': <tensorflow.python.training.tracking.util.Checkpoint at 0x7fefac044a20>,
 '_feature_extractor': <object_detection.models.ssd_resnet_v1_fpn_keras_feature_extractor.SSDResNet50V1FpnKerasFeatureExtractor at 0x7fefac0240b8>

i dont see these on the output above. 3. Third issue is, In exercise 10, loss is not converging to less than 1
Start fine-tuning! batch 0 of 100, loss=1.7889742 batch 10 of 100, loss=1.8075064 batch 20 of 100, loss=1.7422005 batch 30 of 100, loss=1.692342 batch 40 of 100, loss=1.6244757 batch 50 of 100, loss=1.5653871 batch 60 of 100, loss=1.5103792 batch 70 of 100, loss=1.4928862 batch 80 of 100, loss=1.4382222 batch 90 of 100, loss=1.4351206 Done fine-tuning!
4. On exercise 11, on the final test. i get the below output
237 (4,) False False False

and 0.0 on the zombie_detected.

5. and finally, When i submit for grading, obviously it doesnt get pass. comes up with the below message.
   Only 9.28% of the generated boxes match the ground truth boxes with a relative tolerance of 0.3. At least 70.00% are required to pass. please let me know what ami doing wrong.

thanks
Nagarajan.

[gent.spah]

It’s been a long time since I went through that Lab, but what I can say is make sure you don’t change any of the imported libraries to a different one.

Second, it doesn’t seem you are choosing the right layers. I would recheck the steps from exercise 5 or even above it to make sure you are downloading the right model and weights!

[nksubramanian]

Hi, after many tries, my loss is converging now below 1. but still i get the below On exercise 11, on the final test. i get the below output
237 (4,) False False False

and 0.0 on the zombie_detected.

i have submitted my data for grading and again failed. really need someone’s assistance please. i have run out of options and spent nearly 2 days on this.

many thanks
nagarajan,

[gent.spah]

Please send me you ipynb notebook on private message, I will have a look at it!

[Deepti_Prasad]

a score of 0.0 is pointing you to check if your checkpoint code is correct or not

Refer this post, to cross check with your codes

Loss error in exercise 10

Hello Joel, Review of your updated notebook. No corrections done for the points I mentioned in the above comments The only correction you did is where I gave you the direct codes but somehow that also was done incorrectly(probably I didn’t give you the complete solution for you to work on) if v.name.startswith(‘WeightSharedConvolutionalBoxPredictor/WeightSharedConvolutionalBoxHead’) or v.name.startswith(‘WeightSharedConvolutionalBoxPredictor/WeightSharedConvolutionalClassHead’): to_fi…

In case, issue still persist, then need to check the tf and keras version issue.

for checkpoint issue, refer the below linked post comment

C3W2 Assignment Zombie Detector

Hi @hj320 Your checkpoint_path is incorrect checkpoint_path = “models/research/object_detection/test_data/ckpt-33264” hints or instruction on how to recall the checkpoint_path, instruction mentions ckpt-0 but you are using ckpt-33264 but clearly your recall requires more correction, check the below image carefully, a checkpoint would first look for the content then model. Also you missing the point that test_data would have checkpoint file which would contain all the ckpts!!!(I LITERALLY HA…

Regards
DP

[nksubramanian]

Hi Gent, sent you my notebook code through private message. Hope you have received it. thanks.